Regional Climate Model Assessment of the Urban Land-Surface Forcing Over Central Europe

Atmos. Chem. Phys., 14, 12393–12413, 2014 www.atmos-chem-phys.net/14/12393/2014/ doi:10.5194/acp-14-12393-2014 © Author(s) 2014. CC Attribution 3.0 License.

Regional climate model assessment of the urban land-surface forcing over central Europe

P. Huszar1, T. Halenka1, M. Belda1, M. Zak1, K. Sindelarova1,*, and J. Miksovsky1 1Department of Meteorology and Environment Protection, Faculty of Mathematics and Physics, Charles University, Prague, V Holešovickáchˇ 2, Prague 8, 18000, Czech Republic *now at: UPMC Univ. Paris 06; Université Versailles St-Quentin; CNRS/INSU; LATMOS-IPSL, Paris, France

Correspondence to: P. Huszar ([email protected])

Received: 19 May 2014 – Published in Atmos. Chem. Phys. Discuss.: 14 July 2014 Revised: 15 September 2014 – Accepted: 21 October 2014 – Published: 26 November 2014

Abstract. For the purpose of qualifying and quantifying the Statistically significant impacts are modelled not only over climate impact of cities and urban surfaces in general on cli- large urbanized areas, but the influence of the cities is also mate of central Europe, the surface parameterization in re- evident over rural areas without major urban surfaces. It is gional climate model RegCM4 has been extended with the shown that this is the result of the combined effect of the Single-layer Urban Canopy Model (SLUCM). A set of ex- distant influence of the cities and the influence of the minor periments was performed over the period of 2005–2009 for local urban surface coverage. central Europe, either without considering urban surfaces or with the SLUCM treatment. Results show a statistically significant impact of urbanized surfaces on temperature (up to 1.5 K increase in summer) as well as on the boundary layer 1 Introduction height (increases up to 50 m). Urbanization further influences surface wind with a winter decrease up to −0.6 m s−1, though The artificial urban surfaces are clearly distinguished from both increases and decreases were detected in summer de- natural surfaces by mechanical, radiative, thermal and hy- pending on the location relative to the cities and daytime draulic properties. Therefore, these surfaces represent addi- (changes up to 0.3 m s−1). Urban surfaces significantly re- tional sinks and sources of momentum, heat and moisture af- duce the humidity over the surface. This impacts the sim- fecting the mechanical, thermodynamical, and hydrological ulated summer precipitation rate, showing a decrease over properties of local atmosphere and have specific impact on cities of up to −2 mm day−1. Significant temperature in- the meteorological conditions (Oke, 1982, 1987; Godowitch creases are simulated over higher altitudes as well, not only et al., 1985; Eliasson and Holmer, 1990; Haeger-Eugensson within the urban canopy layer. With the urban parameteriza- and Holmer, 1999). tion, the climate model better describes the diurnal tempera- One of the most comprehensively studied aspects of the ture variation, reducing the cold afternoon and evening bias meteorological impact of urban surfaces is the urban heat is- of RegCM4. land (UHI) phenomenon, which represents an excess warmth Sensitivity experiments were carried out to quantify the of urbanized areas with respect to their non-urbanized (rural) response of the meteorological conditions to changes in vicinity. In general, UHI forms due to significant perturba- the parameters specific to the urban environment, such as tion of fluxes of energy, moisture and momentum within this street width, building height, albedo of the roofs and anthro- environment, which is characterized by canyon-like geome- pogenic heat release. The results proved to be rather robust try and specific thermal parameters of the artificial surfaces and the choice of the key SLUCM parameters impacts them (Oke, 1982). Due to their decreased albedo, urban surfaces only slightly (mainly temperature, boundary layer height and store more heat compared to rural areas and after sunset this wind velocity). heat is released with a reduced efficiency because of the decreased sky-view factor (Grimmond and Oke, 1995) mak-

Published by Copernicus Publications on behalf of the European Geosciences Union. 12394 P. Huszar et al.: Urban land-surface impact on climate ing UHI typical for night-time, although it is detectable dur- of local urban features on the mesoscale meteorology and ing daytime as well. The UHI is enhanced by several other climate, despite their increasing resolution. Therefore the in- factors. Increased anthropogenic heat emission within ur- clusion of urban canopy models (UCMs), which are specially ban environment increases urban temperatures (Block et al., designed to parameterize the processes specific to the urban 2004). Further, urban areas covered by impervious surfaces environment that are not resolvable at the model’s scale, is exhibit higher runoff than their rural counterparts, which necessary (Baklanov et al., 2008; Lee and Park, 2008; Ole- leaves them with less surface water available for evapora- son et al., 2008; Chen et al., 2011). tion. Lower evaporation decreases the latent heat consump- Along with the development of UCMs, many modelling tion causing the perturbation of energy balance, which leads studies were carried out to describe UHI and other aspects to temperature increase (Grimmond and Oke, 1991; Taha, of the urban-atmosphere interaction. Most of the studies fo- 1997). cused on a particular city with minor interest in the impact Some studies performing site measurement in and around on regional scale further from the urban area itself. Exam- cities revealed the so-called urban cool island (UCI) effect ples of such recent modelling studies include the following: as well: during the morning hours, the enhanced shadowing Klaic´ et al.(2002) for Zagreb, Croatia; Flagg and Taylor within urban surfaces delays the heating and causes lower (2011) for Detroit-Windsor, USA; Giannaros et al.(2013) temperatures than over the rural surfaces (Basara et al., 2008; for Athens, Greece; Wouters et al.(2013) for Paris, France; Gaffin et al., 2008). Miao et al.(2009) and Hou et al.(2013) for the Beijing area, Urban surfaces have further impacts on other meteoro- China. These studies conducted experiments for time peri- logical parameters as well: Richards and Oke(2002) and ods of several days during selected weather episodes only. Richards(2004) studied the changes of surface humidity, Furthermore, as mentioned earlier, they focused on the scales while Grimmond and Oke(2002), Roth(2000) or Kastner- of at most a few tens of kilometres around a particular city. Klein et al.(2001) focused on the impact on roughness Recently, Zhang et al.(2009), Wang et al.(2012) and Feng and turbulence. Many studies dealt with the structure of the et al.(2013) examined the impact of urban land surfaces on urban boundary layer including the impact on the height the regional climate but with focus on eastern Asia. Over of the planetary boundary layer (ZPBL) (Piringer, 2001; Europe, Struzewska and Kaminski(2012) examined the re- Cleugh and Grimmond, 2001; Martilli, 2002; Angevine et al., gional impact of urban surfaces on the forecast of local and 2003; Nair et al., 2004) and wind speed (Hou et al., 2013). regional meteorological conditions and selected pollutants. Urbanization-triggered changes in precipitation and hydro- However, they were not interested in the long-term impacts logical processes also got the attention of research studies and performed simulations of only a few tens of hours for (Dettwiller and Changon, 1976; Shepherd et al., 2002; Ro- three selected episodes. Block et al.(2004) carried out ex- zoff et al., 2003). There is evidence that the urban environ- periments on a regional scale over central Europe as well, but ment with its higher air pollution is responsible for enhanced they were interested in the impact of the anthropogenic heat lightning (Farias et al., 2009; Coquillat et al., 2013). Schal- release (AHR) only, without any attention to other aspects of dach and Alcamo(2007) showed significant influence on the urban climate interactions (e.g. radiation-induced UHI phe- carbon balance as well and, finally, the urban-meteorology nomenon). interaction may significantly influence air quality (Rigby and The study performed by Trusilova et al.(2008) over Eu- Toumi, 2008; Ryu et al., 2013a,b). However, most of the rope used model MM5 to examine the impact of urban sur- influences listed here have to be viewed in a common UHI- faces on climate in the months of July and January over 5 related framework as they are all physically connected within consecutive years. They used a single-layer parameterization this phenomenon, bringing higher street level temperatures with simplified urban geometry representation (TEB – Town and having a direct impact on human health (Reid et al., Energy Budget model; Masson, 2000). 2009) and, in general, on the comfort of living. The present study is one of the first that aims to examine Numerous studies aimed to find observational (surface the regional and long-term impact of all urban surfaces on measurements, reanalysis or satellite based) evidence for the the regional climate of central Europe. To achieve this goal, UHI in different parts of the Earth in the past (e.g. Eliasson the regional climate model RegCM4.2 was coupled with the and Holmer, 1990; Basara et al., 2008; Svoma and Brazel, Single-layer Urban Canopy Model (SLUCM) that accounts 2010; Yang et al., 2011; Zhou and Reng, 2011; Giannaros for the most relevant processes specific to the urban environ- and Melas, 2012; Pichierri et al., 2012). To provide a re- ment including the AHR. The climate impact of urbanization liable, numerical modelling based perspective of the UHI for day and night-time conditions is examined separately, for phenomenon, and of other related impacts (e.g. on wind winter and summer months. The results are evaluated against speed/direction, precipitation, ZPBL), the complex nature surface measurements as well. A few sensitivity experiments of the mechanical, thermodynamical and radiative processes are carried out to examine the results based on the setting of have to be realistically represented in models. Current op- the key parameters of the SLUCM and on different resolu- erational numerical weather prediction models as well as re- tions of the surface model. Finally, it is shown how a partic- gional climate models still fail to capture properly the impact

Atmos. Chem. Phys., 14, 12393–12413, 2014 www.atmos-chem-phys.net/14/12393/2014/ P. Huszar et al.: Urban land-surface impact on climate 12395 ular city (Prague) influences the region by disregarding all of soil water. Within BATS, 20 vegetation types are distin- urban surfaces except those within this city. guished. An improvement in describing the land surface fluxes within the BATS scheme can be achieved by a sub-grid land 2 Tools and experimental set-up surface configuration by which each model grid point is di- vided into a regular sub-grid, and land surface processes are 2.1 The regional climate model RegCM4.2 calculated at each sub-grid point taking into account the local land use and topography (Giorgi et al., 2003). The regional climate model used in this study is the model To describe urbanized surfaces, two new land use types RegCM version 4.2 (hereafter referred to as RegCM4.2), a 3- were added to BATS with the introduction of RegCM ver- D mesoscale model developed at The International Centre sion 4.0 and higher. These new land use types alter the values for Theoretical Physics (ICTP) and is described in full by of albedo, roughness length, soil characteristics, and maxi- Giorgi et al.(2012). Its dynamical core is based on the hy- mum vegetation cover in order to account for the modified drostatic version of the NCAR-PSU Mesoscale Model ver- surface energy balance (heat and momentum), evapotranspi- sion 5 (MM5) (Grell et al., 1994). The radiation is solved ration and runoff specific to urbanized surfaces (parameters using the Community Climate Model version 3 (CCM3) are taken from Kueppers et al., 2008). This represents a bulk parameterization (Kiehl et al., 1996). The large-scale pre- parameterization of zero-order effects of urban and suburban cipitation and cloud processes are calculated following Pal land use types, which however ignores the 3-D character of et al.(2000) and the convection is parameterized with ei- the processes that occur in the urban environment (e.g. in ther the the Grell scheme (Grell, 1993), Kuo scheme (Anthes, street-canyons). Therefore, we implemented a more sophis- 1977) or the MIT-Emanuel scheme (Emanuel and Zivkovic- ticated treatment of the meteorological processes that occur Rothman, 1999). The planetary boundary layer scheme is in connection with urban surfaces. Chen et al.(2011) pro- based on Holstag et al.(1990). vided an overview of the urban parameterizations that are RegCM4.2 includes two land surface models: Bio- implemented in the Weather Research and Forecasting model sphere–Atmosphere Transfer Scheme (BATS), originally de- (WRF; Skamarock et al., 2008). For our study, the Single- veloped by Dickinson et al.(1993); and the Community Land layer Urban Canopy Model (SLUCM) is used. It is less Surface Model v3.5 (CLM3.5) (Oleson et al., 2008; Tawfik complex compared to the multi-layer urban canopy models and Steiner, 2011) as an optional land surface parameteriza- (MLUCMs) and therefore computationally less demanding tion. in long term climate model simulations. On the other hand, it The BATS scheme is designed to describe the role of accounts for the 3-D meteorological processes occurring in vegetation and interactive soil moisture in modifying the the cities’ environment, such as trapping the radiation within surface-atmosphere exchanges of momentum, energy, and the street-canyon and shadowing due to buildings. The fol- water vapour. It considers a vegetation layer, a snow layer, lowing section gives a more detailed description of SLUCM. a surface soil layer (10 cm thick) or root zone layer (1–2 m thick) and a deep soil layer (3 m thick). Soil layer tempera- 2.2 The Single-layer Urban Canopy Model tures are calculated using a generalization of the force-restore method of Deardoff(1978). The energy balance formulation The Single-layer Urban Canopy Model was developed by (including sensible, radiative, and latent heat fluxes) is used Kusaka et al.(2001) and Kusaka and Kimura(2004). It rep- for computing temperature of the canopy and canopy foliage. resents the geometry of cities assuming infinitely long street The soil hydrology calculations include predictive equa- canyons, where it considers shadowing, reflections, and trap- tions for the water content of the soil layers, accounting for ping of radiation and prescribes an exponential wind profile precipitation, snowmelt, canopy foliage drip, evapotranspira- (Fig.1). The model calculates the surface skin temperatures tion, surface runoff, infiltration below the root zone, and dif- of the roof, wall, and road (determined from the surface en- fusive exchange of water between soil layers. Surface runoff ergy budget) and temperature profiles within roof, wall, and rates are expressed as functions of the precipitation and soil road layers (determined from the thermal conduction equa- water saturation. Snow depth is prognostically calculated tion) as prognostic variables. Surface-sensible heat fluxes from snowfall, snowmelt, and sublimation. It is assumed that from each surface are calculated using Monin–Obukhov sim- precipitation falls in the form of snow if the temperature of ilarity theory and the Jurges formula, which is widely used the lowest model level is below 271 K. in the field of architecture (e.g. Tanaka et al., 1993). SLUCM Sensible heat, water vapour, and momentum fluxes at the calculates canyon drag coefficient and friction velocity us- surface are calculated using a standard surface drag coeffi- ing a similarity stability function for momentum. For a de- cient formulation based on surface-layer similarity theory. tailed description of the SLUCM, please refer to Kusaka et al. The drag coefficient depends on the surface roughness length (2001). and on the atmospheric stability in the surface layer. The The implementation of SLUCM into RegCM4.2 follows surface evapotranspiration rates depend on the availability the way of its coupling to the WRF model (Chen et al., 2011). www.atmos-chem-phys.net/14/12393/2014/ Atmos. Chem. Phys., 14, 12393–12413, 2014 412396 P. HuszarP. Huszar et al.: et al.: Urban Urban land-surface land-surface impact impact on climate on climate

tion (e.g. western Ukraine, Belarus), the Global Land Cover 2000 (GLC, 2000) database is used. In particular, urban category is a compilation of continuous urban fabric land use type (cat. 1.1.1 from Corine2006) and artificial surface and associated areas (cat. 22 from GLC2000). Sub-urban category is identical to discontinuous urban fabric land use type (cat. 1.1.2 from Corine2006). Fig- ure2 presents the model domain with orography and urban land coverage. Both land use types (urban and sub-urban) are characterized by specific urban geometry parameters (Table1). The urban canopy parameters used in the SLUCM are the ones used in the WRF implementation of SLUCM (see Chen Figure 1. A schematic representation of the SLUCM (after Chen Figureet al., 1.2011A). schematic representation of the SLUCM (after Chen et al., 2011) modified to better describe the urban environ- et al., 2011). ment in cities in central Europe. Building heights were increased and are similar to those used for Polish cities by StruzewskaFigure 2. and Kaminski(2012). Similar values were used SLUCM is coupled to the RegCM4.2 within the BATS sur- The model domain with orography in m at 10km 10km and the urban canyons is then passed to the RegCM4.2’s by resolutionMartilli(2002 and) urban who coverage focused onat European citiesresolution as well. (fine-scale× face model, eventually using the BATS’s subgrid treatment 2km 2km Thebrown road/street elements). widths are also more suited× to the targeted BATS(SUBBATS). model. The Whenever total momentum BATS finds a flux subgrid is passed box covered back in a similar way. BATS then calculates the overall flux for the region and follow the values of Martilli(2002) or Ratti et al. by urban surface in the subgrid land use file, it calls SLUCM (2001). The values for the AHR represent the annual average modelroutines. grid The box total by aggregating sensible heat the flux fluxes fromfrom roofs, non-urban walls, roads, and whileurban the fabric model land internally use type calculates (cat. 1.1.2 the frommonthly Corine2006). variation Fig- 280 urbanandsurfaces the urban provided canyons by is SLUCM. then passed Similarly, to the RegCM4.2’s the total fric- (theure yearly 2 presents amplitude the is model about domain 70 to 130 with % of orography the annual and av- urban tionBATS velocity model. is The aggregated total momentum from urban flux and is passed non-urban back insur- erage value for July to January, respectively). They follow the 315 land coverage. facesa similar and passed way. BATS to RegCM4.2 then calculates boundary-layer the overall scheme.flux for the The case of a Polish city (Kłysik, 1995) but are also in line with anthropogenicmodel grid box heat by aggregating release (AHR) the fluxes and from its diurnal non-urban variation and the measurementsBoth land use from types the (urbancity of Toulouse, and sub-urban) France (arePigeon character- areurban considered surfaces and provided added by to SLUCM. the sensible Similarly, heat the flux total from fric- the etized al., 2007 by specific), with the urban consideration geometry that parameters the central (Table Euro- 1). The 285 urbantion canopy velocity layer. is aggregated from urban and non-urban sur- peanurban climate canopy during parameters winter is usedcolder, in bringing the SLUCM higher are heat- the ones faces and passed to RegCM4.2 boundary-layer scheme. The ingused and largerin the heat WRF release. implementation Other parameters of SLUCM of the urban (see Chen 2.3anthropogenic Model configuration heat release and (AHR) experiments and its diurnal performed variation 320canopyet al., in 2011) SLUCM modified are unchanged to better from describe the WRF the implemen- urban environ- are considered and added to the sensible heat flux from the tation.ment It inhas cities to be noted in central that these Europe. values Building are rough heights estimates were in- Severalurban canopy experiments layer. are performed using the andcreased describe and all urban are similar (and suburban) to those surfaces used for within Polish the do- cities by RegCM4.2/SLUCM model over 10km 10km resolu- main.Struzewska However, and there Kaminski are large (2012). differences Similar between values cities, were used tion2.3 domain Model configuration of 158 118 andgridpoints experiments× covering performed central andby even Martilli within (2002) our 2km who× focused2km subgrid on European box, these cities values as well. × 290 Europe. The initial and lateral boundary conditions for 325canThe vary road/street substantially. widths Because are of also this, more a possible suited uncertainty to the targeted RegCM4.2Several were experiments taken from are the performed ERA Interim using data (Sim- the is broughtregion and into follow the results. the values To assess of Martilli its magnitude, (2002) a or sensi- Ratti et al. × monsRegCM4.2/SLUCM et al., 2007). The model convection over 10km was parameterized10km resolu- with tivity(2001). test to The the values key urban for the parameters AHR represent was included the annual in this average tion domain of 158 × 118 grid points covering central study (Sect. 3.3). the Grell scheme with the closure assumption of Fritsch while the model internally calculates the monthly variation Europe. The initial and lateral boundary conditions for Table 2 summarizes the experiments performed using the and Chappell (1980). The BATS scheme was configured (the yearly amplitude is about 70 % to 130 % of the annual RegCM4.2 were taken from the ERA-Interim data (Sim- RegCM4.2 extended with the SLUCM. The NOURBAN ex- 295 with a 2km 2km subgrid set-up, which provides the 330 average value for July to January, respectively). They follow mons et al., ×2007). The convection was parameterized with periment corresponds to the case with no urban coverage, opportunitythe Grell scheme to resolve with and the to closure describe assumption medium ofand Fritsch large i.e.the the case urban of land a Polish use types city were (Kłysik, changed 1995) to but the are dominat- also in line continuousand Chappell urban(1980 surfaces). The BATS (cities) scheme over the was targeted configured area. ingwith adjacent the measurements land use categories, from usually the city “crop” of Toulouse, and “for- France Awith further a increase2km × 2 ofkm thesubgrid subgrid set-up, division which would provides be definitely the est”,(Pigeon similarly et as al., in 2007),Trusilova with et al. the(2008 consideration). The SLUCM that ex- the cen- beneficial,opportunity but to it comesresolve with and a to significant describe medium reduction and of large model perimentstral European considers climate urban during surfaces, winter treating is colder, them bringing with the higher 300 speed.continuous urban surfaces (cities) over the targeted area. 335SLUCMheating urban and canopy larger heat parameterization. release. Other The parameters mentioned ofex- the ur- ASLUCM further increasewas configured of the subgrid using four division surface would layers definitely of roof perimentsban canopy were in run SLUCM for a 5-year are unchanged period between from 2005–2009. the WRF imple- (eachbe beneficial, 5 cm thick), but wallit comes (each with 5 cm a significantthick) and reduction road (5 ofcm, Thementation. impact of It urban has tosurfaces be noted on climate that these was values then evaluated are rough esti- 25modelcm, 50 speed.cm and 75 cm). The inner building temperature asmates the difference and describe between all SLUCM urban (and and suburban) NOURBAN surfaces for the within wasSLUCM set to a constant was configured value of using 298 fourK. surface layers of roof selectedthe domain. meteorological However, fields. there are large differences between (each 5 cm thick), wall (each 5 cm thick) and road (5, 25, To analyse the sensitivity of the results to the key urban 305 Land use information was compiled using Corine2006 340 cities, and even within our 2km 2km subgrid-box, these (EEA,50 and 2012) 75 cm). dataset The andinner where building it does temperature not provide wasset informa- to a parameters,values can four vary additional substantially. simulations Because× were performed of this, witha possible constant value of 298 K. modifications of the building height, street width, roof albedo tion (e.g. western Ukraine, Belarus), the Global Land Cover uncertainty is brought into the results. To assess its mag- Land use information was compiled using Corine2006 and AHR (Table2). These simulations were performed for a 2000(EEA (GLC,, 2012 2000)) data setdatabase and where is used. it does not provide informa- 1-yearnitude, period a sensitivity (2005) and aretest titled to the as SEN1–4. key urban parameters was In particular, urban category is a compilation of Continu- included in this study (Sect. 3.3). Table 2 summarizes the

310 ous urban fabric land use type (cat. 1.1.1 from Corine2006) 345 experiments performed using the RegCM4.2 extended with andAtmos. Artiﬁcial Chem. surface Phys., 14,and 12393– associated12413, areas 2014 (cat. 22 from the SLUCM. www.atmos-chem-phys.net/14/12393/2014/ The NOURBAN experiment corresponds to the GLC2000). Sub-urban category is identical to Discontinuous case with no urban coverage, i.e. the urban landuse types 4 P. Huszar et al.: Urban land-surface impact on climate P. Huszar et al.: Urban land-surface impact on climate 12397

Figure 1. A schematic representation of the SLUCM (after Chen et al., 2011).

Figure 2. The model domain with orography in m at 10km × 10km resolution and urban coverage at 2km × 2km resolution (fine-scale brownFigure elements). 2. The model domain with orography in m at 10km 10km × and the urban canyons is then passed to the RegCM4.2’s resolution and urban coverage at 2km 2km resolution (fine-scale In the noSUBBATS experiment, it is examined how the impact× are presented separately. Furthermore, as the meteo- BATS model. The total momentum flux is passed back in resultsbrown change elements). if the surface model resolution equals to rological regime in the urban environment differs between the dynamical resolution (i.e. 10 km). As a reference sim- day and night-time, we will discuss separately the day and a similar way. BATS then calculates the overall flux for the ulation, a similar experiment is performed without SUB- night impacts as well. Shaded areas in all “spatial” figures BATS and without any urban treatment (denoted noSUB- represent statistically significant differences at the 95 % level model grid box by aggregating the fluxes from non-urban and BATS/NOURBAN). Finally, to see the extent of the urban according to the one sample t test of the difference of two surfaces’ influence on remote areas, an experiment titled fields (for variables with approximately Gaussian distribu- 280 urban surfaces provided by SLUCM. Similarly, the total fric- PRAGUEurban was fabric performed, land where only use the citytype of Prague (cat. was 1.1.2tion of from the differences), Corine2006). except for wind speed Fig- and precipita- treated as urban surface. Here, Prague is defined as a squared tion, where the non-parametrical sign-test was applied. tion velocity is aggregated from urban and non-urban sur- regionure of 2 50km presents× 50km area the centred model on the city’s domain midpoint. withFigure3 presents orography the day (upper and panels) urban and night-time RegCM4.2 was run for the year 2004 without SLUCM as (bottom panels) impacts of urban surfaces on the 2 m air tem- faces and passed to RegCM4.2 boundary-layer scheme. The 315 aland spin-up coverage. time, and all the experiments were restarted from perature averaged over winter (left) and summer (right). In this run. general, the impact is highest over highly urbanized areas in- anthropogenic heat release (AHR) and its diurnal variation Both land use types (urban anddicating sub-urban) a well pronounced are UHI effect. character- The location of cities such as Berlin, Prague, Vienna, Budapest, Munich or War- are considered and added to the sensible heat flux from the 3ized Results by specific urban geometrysaw parameters is well recognized through (Table the UHI 1). effect. The In winter, the impact is generally lower, reaching values up to 0.4 K over 285 urban canopy layer. 3.1urban Spatial canopy distribution of parameters the impact of urban surfacesused indaytime the for SLUCM many cities. However, are the the impact ones is statistically significant even over rural regions far from larger cities, with Thisused section in presents the the spatial WRF (horizontal implementation and vertical) dis- up to 0.05 of K SLUCM temperature increase. (see During Chen night, the impact tribution of the meteorological impacts of urbanized sur- is very small and rather noisy, though there is an indication 2.3 Model configuration and experiments performed 320 faces.et al., To evaluate 2011) their modifiedseasonal dependence, to the better winter describe the urban environ- (DJF;ment December–February) in cities and in summer central (JJA; June–August) Europe. Building heights were in- Several experiments are performed using the www.atmos-chem-phys.net/14/12393/2014/creased and are similar to those used for Atmos. Polish Chem. Phys., cities 14, 12393– by12413, 2014 RegCM4.2/SLUCM model over 10km 10km resolu- Struzewska and Kaminski (2012). Similar values were used tion domain of 158 118 gridpoints× covering central by Martilli (2002) who focused on European cities as well. × 290 Europe. The initial and lateral boundary conditions for 325 The road/street widths are also more suited to the targeted RegCM4.2 were taken from the ERA Interim data (Sim- region and follow the values of Martilli (2002) or Ratti et al. mons et al., 2007). The convection was parameterized with (2001). The values for the AHR represent the annual average the Grell scheme with the closure assumption of Fritsch while the model internally calculates the monthly variation and Chappell (1980). The BATS scheme was configured (the yearly amplitude is about 70 % to 130 % of the annual 295 with a 2km 2km subgrid set-up, which provides the 330 average value for July to January, respectively). They follow opportunity to× resolve and to describe medium and large the case of a Polish city (Kłysik, 1995) but are also in line continuous urban surfaces (cities) over the targeted area. with the measurements from the city of Toulouse, France A further increase of the subgrid division would be definitely (Pigeon et al., 2007), with the consideration that the cen- beneficial, but it comes with a significant reduction of model tral European climate during winter is colder, bringing higher

300 speed. 335 heating and larger heat release. Other parameters of the ur- SLUCM was conﬁgured using four surface layers of roof ban canopy in SLUCM are unchanged from the WRF imple- (each 5 cm thick), wall (each 5 cm thick) and road (5 cm, mentation. It has to be noted that these values are rough esti- 25 cm, 50 cm and 75 cm). The inner building temperature mates and describe all urban (and suburban) surfaces within was set to a constant value of 298 K. the domain. However, there are large differences between 305 Land use information was compiled using Corine2006 340 cities, and even within our 2km 2km subgrid-box, these (EEA, 2012) dataset and where it does not provide informa- values can vary substantially. Because× of this, a possible tion (e.g. western Ukraine, Belarus), the Global Land Cover uncertainty is brought into the results. To assess its mag- 2000 (GLC, 2000) database is used. nitude, a sensitivity test to the key urban parameters was In particular, urban category is a compilation of Continu- included in this study (Sect. 3.3). Table 2 summarizes the

310 ous urban fabric land use type (cat. 1.1.1 from Corine2006) 345 experiments performed using the RegCM4.2 extended with and Artiﬁcial surface and associated areas (cat. 22 from the SLUCM. The NOURBAN experiment corresponds to the GLC2000). Sub-urban category is identical to Discontinuous case with no urban coverage, i.e. the urban landuse types 12398 P. Huszar et al.: Urban land-surface impact on climate

Table 1. Urban canopy parameters of SLUCM as implemented in RegCM4.2 for the central European region.

Parameter Unit Urban Sub-urban h (building height) m 20 15 lroof (roof width) m 20 15 lroad (street width) m 15 20 AH (anthropogenic heat) Wm−2 70 30 Furb (urban fraction) – 0.9 0.7 −3 −1 6 6 CR (heat capacity of the roof) Jm K 1.0 × 10 1.0 × 10 −3 −1 6 6 CW (heat capacity of the wall) Jm K 1.0 × 10 1.0 × 10 −3 −1 6 6 CG (heat capacity of the road) Jm K 1.4 × 10 1.4 × 10 −1 −1 −1 λR (thermal conductivity of the roof) Jm s K 0.67 0.67 −1 −1 −1 λW (thermal conductivity of the wall) Jm s K 0.67 0.67 −1 −1 −1 λG (thermal conductivity of the road) Jm s K 0.40 0.40 αR (albedo of the roof) – 0.20 0.20 αW (albedo of the wall) – 0.20 0.20 αG (albedo of the road) – 0.20 0.20 εR (emissivity of the roof) – 0.90 0.90 εW (emissivity of the wall) – 0.90 0.90 εG (emissivity of the road) – 0.95 0.95 Z0R (roughness length for momentum over roof) m 0.01 0.01

Table 2. Different experiments performed with the RegCM4.2/SLUCM couple. The sensitivity runs are described in terms of the percentage change of the selected urban parameter with respect to the value listed in Table1.

Experiment Urban treatment Period SLUCM parameters SUBBATS Urban surfaces NOURBAN No 2005–2009 – Yes – SLUCM Yes 2005–2009 Default – see Table1 Yes All SEN1 Yes 2005 50 % building height Yes All SEN2 Yes 2005 50 % road width Yes All SEN3 Yes 2005 200 % roof albedo Yes All SEN4 Yes 2005 50 % AHR Yes All noSUBBATS Yes 2005–2009 Default – see Table1 No All noSUBBATS/NOURBAN No 2005–2009 – No – PRAGUE Yes 2005–2009 Default – see Table1 Yes Prague only for a slight temperature decrease over cities up to −0.1 K with significant ZPBL increase (up to 50 m over most of the (over Berlin, Prague, Vienna, Budapest and others). domain). During winter night-time, just a limited number of A much stronger impact is modelled during summer. Dur- locations exhibit a significant change, with up to −20 m de- ing the day, temperature increased almost everywhere over crease. In summer, extensive areas encounter ZPBL increase the domain, with a peak over the cities less pronounced than due to urbanized surfaces (up to 100 m), not limited to large during night-time. The impact exceeds 0.4 K over a large cities only. During night-time, the impact of cities is much part of the domain far from urbanized centres and is high- smaller and is characterized by an increase of ZPBL (up to est over Budapest (0.7 K) and Milan (1.0 K). As expected, 50 m). However, over rural areas, a statistically significant the most pronounced impact is modelled during the summer decrease is modelled (up to −20 m). night-time when the UHI phenomenon is considered to be The impact on 10 m wind velocity (Fig.5) depends largely the strongest. Over many cities, the temperature increase ex- on whether winter or summer conditions and day or night ceeds 1.5 K (Berlin, Prague, Vienna, Budapest, Munich, Mi- are considered. In winter daytime, urban effects can result in lan, etc.), and is also substantial over rural areas, reaching both small increase (usually in the vicinity of the cities) and 0.2 K. small decrease of wind speed (located over cities). A more The changes in the height of the planetary boundary layer uniform picture is visible for the night-time changes in win- (ZPBL) are presented in Fig.4. The daytime impacts are usu- ter: there is an evident wind speed reduction especially over ally more prominent, indicating an increase up to 50 m in the cities up to −0.6 ms−1. The decrease over rural areas winter and up to 200 m in the summer months. In winter, the is small (−0.1 ms−1) and usually statistically insignificant. statistically significant changes are usually limited to larger The impact of cities on wind velocity in summer is charac- cities, while in summer, large rural areas are also marked terized, during daytime, by a decrease just over the cities (up

Atmos. Chem. Phys., 14, 12393–12413, 2014 www.atmos-chem-phys.net/14/12393/2014/ 6 P. Huszar et al.: Urban land-surface impact on climate

P. Huszar et al.: Urban land-surface impact on climate 12399

Figure 3. The impact of urbanized surfaces on the winter (left) and summer (right) near surface temperature for day (above) and night-time Figure 3. The impact(bottom) of urbanized conditions surfaces in K averaged on over the the winter years 2005–2009. (left) and Shaded summer areas represent (right) statistically near significant surface differences temperature on the 95 for % level. day (above) and nighttime (bottom) conditions in K averaged over years 2005–2009. Shaded areas represent statistically significant differences on the 95 % level. to −0.2 ms−1), with a small but statistically significant in- −2 mmday−1, mostly over Budapest, Katowice, Warsaw, Vi- crease just around the cities (up to 0.2 ms−1). During night- enna, Prague, Munich and the Ruhr region in Germany. The time, urban surfaces seem to increase the wind speed up to precipitation changes over areas without dominant urban sur- 0.3 ms−1, although this is not evident for all major urban faces are usually statistically insignificant. Near a few cities, centres throughout central Europe, but rather limited to cities a slight but statistically significant increase of precipitation over the western part of the domain. is modelled as well, mainly in central Germany and Poland, As the variables related to water budget (evaporation, hu- up to 0.5 mmday−1. midity, precipitation) were influenced significantly only dur- Apart from the urban induced changes to the horizontal ing summer, we present results only for this season. De- distribution of meteorological fields, perturbations to the ver- crease of evaporation commonly associated with urban sur- tical structure of temperature are presented here. Figure7 faces comes along with reduced specific humidity that can be illustrates the temperature vertical cross-section along the well seen in Fig.6 (left). During summer, especially for day- 50◦ N latitude, which crosses several urbanized areas. A sta- time, the decrease of specific humidity can exceed −1 gkg−1 tistically significant warming is indicated up to about 1.7 km over cities (peaking at −1.3 gkg−1). During night, the de- in summer, i.e. approximately in the planetary boundary crease is of smaller magnitude, up to −0.6 gkg−1. The de- layer (PBL). The warming is most intense just above the sur- crease is not limited to larger central European cities only, face, but remains above 0.2 K almost across the entire PBL but it affects almost the entire domain for both day and night. depth during both day and night. The warming due to urban The introduction of urban surfaces has an impact on the surfaces spreads to lower elevations (300 m) in winter, due precipitation rates in our simulations as well (Fig.6, left). to reduced vertical mixing, and is less significant during the The effect is statistically significant during summer day- night. Above the boundary layer in summer (between alti- time in connection with the decreased evaporation, when it tudes 1.7 and 10 km), statistically significant cooling is mod- is characterized with a large decrease over cities exceeding elled, especially during daytime (up to −0.1 K). Finally there

www.atmos-chem-phys.net/14/12393/2014/ Atmos. Chem. Phys., 14, 12393–12413, 2014

Figure 4. Same as Fig. 3 but for the PBL height in meters. tion of the differences), except for wind speed and precipita- such as Berlin, Prague, Vienna, Budapest, Munich or War- tion, where the non-parametrical sign-test was applied. 395 saw is well recognized through the UHI effect. In winter, the Figure 3 presents the day (upper panels) and nighttime impact is generally lower, reaching values up to 0.4 K over 390 (bottom panels) impacts of urban surfaces on the 2 m air tem- daytime for many cities. However, the impact is statistically perature averaged over winter (left) and summer (right). In signiﬁcant even over rural regions far from larger cities with general, the impact is highest over highly urbanized areas in- up to 0.05 K temperature increase. During night, the impact dicating a well pronounced UHI effect. The location of cities 400 is very small and rather noisy, though there is an indication 6 P. Huszar et al.: Urban land-surface impact on climate

Figure 3. The impact of urbanized surfaces on the winter (left) and summer (right) near surface temperature for day (above) and nighttime (bottom) conditions in K averaged over years 2005–2009. Shaded areas represent statistically signiﬁcant differences on the 95 % level.

12400 P. Huszar et al.: Urban land-surface impact on climate

Figure 4. Same as Fig.3 but for the PBL height in metres. Figure 4. Same as Fig. 3 but for the PBL height in meters.

is an indication of warming above approximately 10 km (up mation about the magnitude of the UHI effect, the figure in- to 0.05 K), but mostly statistically insignificant. In winter, no cludes the average temperature from the vicinity of the cities tion of the differences),statistically except significant for wind impact speed is modelled and above precipita- the PBL. (definedsuch as approximately Berlin, Prague, a 10 km wide Vienna, belt 20 km Budapest, from the Munich or War- city centres) taken from the SLUCM experiment. In our sim- tion, where the non-parametrical3.2 Monthly and sign-test daily temperature was applied. profiles and 395 ulations,saw is the well UHI recognized develops during through noon hours the and UHI reaches effect. In winter, the Figure 3 presents thecomparison day (upper with observations panels) and nighttime itsimpact maximum is around generally 10 p.m. to lower, midnight reaching local time at values about up to 0.4 K over 0.5 K on annual average. The UHI diminishes around 6 a.m. 390 (bottom panels) impactsThis section of urban presents surfaces the impact on of the urbanized 2 m air surfaces tem- on daytimeDuring the for morning many hours cities. until noon, However, a weak UCI the (ur- impact is statistically perature averaged overthe monthly winter and (left) hourly and temperature summer variations (right). over In se- bansignificant cool island) even develops over over rural Berlin, regions Budapest andfar Mu- from larger cities with general, the impact islected highest cities in over central highly Europe. urbanized Table3 shows areas the monthly in- nich,up with to 0.05 maximumtemperature temperature decrease increase. of −0.5 to During−0.1 K night, the impact 2 m temperatures averaged over the years 2005–2009 from reaching at aroundK 9–11 a.m. dicating a well pronouncedboth NOURBAN UHI andeffect. SLUCM The simulations location for of the cities summer400 isIt very is interesting small to and see that rather the city noisy, temperatures though for con- there is an indication months (June, July and August). The corresponding values figuration disregarding the urban surfaces (NOURBAN, blue from the E-OBS gridded observational data set (Haylock line) are lower than the temperatures accounting for urban et al., 2008) are included for comparison as well. A system- effect in the vicinity of cities during the maximum UHI de- atic underestimation of temperatures is revealed throughout velopment (SLUCM, orange line), by up to 0.1 K on annual the whole year (up to −4 K over Berlin in spring – not pre- average. This is consistent with our spatial results, which sented here). The table clearly shows, that, during summer showed that the impact of urban surfaces is not limited to months, the negative model bias is eliminated (or at least re- the air column over large cities, but is spread over remote duced) by introducing the SLUCM parameterization for the areas as well, as seen in Fig.3. urban surfaces. This holds especially for Munich and Bu- For more detailed analysis, hourly station data from three dapest. locations in Prague are used to compare the observed hourly The 2005–2009 average hourly variation of 2 m tempera- temperature variations and the modelled summer UHI. One tures for selected cities is presented in Fig.8. To obtain infor-

Atmos. Chem. Phys., 14, 12393–12413, 2014 www.atmos-chem-phys.net/14/12393/2014/ 8 P. Huszar et al.: Urban land-surface impact on climate P. Huszar et al.: Urban land-surface impact on climate 12401

Figure 5. Same as Fig.3 but for the wind velocity at 10 m in ms −1. 1 Figure 5. Same as Fig. 3 but for the wind velocity at 10 m in m s− . Table 3. The mean 2005–2009 monthly summer 2 m temperatures averaged over selected cities in central Europe for the NOURBAN, SLUCM runs and extracted from E-OBS observational data.

Prague Vienna Budapest Munich Berlin NOURBAN SLUCM E-OBS NOURBAN SLUCM E-OBS NOURBAN SLUCM E-OBS NOURBAN SLUCM E-OBS NOURBAN SLUCM E-OBS June 15.8 16.4 18.0 18.6 19.6 21.0 19.6 20.3 21.0 15.7 16.5 19.0 15.9 16.4 18.0 July 18.4 18.9 20.0 21.2 22.1 22.0 22.4 22.9 23.0 18.4 19.1 21.0 18.4 19.0 19.0 August 16.9 17.4 18.0 19.7 20.6 20.0 20.4 21.0 21.0 16.7 17.4 18.0 16.6 17.1 17.0

station lies in the inner centre of Prague, where the maxi- thermore indicated by lower temperatures in the city vicinity mum UHI is expected, while two are located in the vicin- in both the SLUCM experiment (by almost 0.5 K) and in the ity of the city centre (about 10–15 km far). The results measurements (up to 1 K). When urban surfaces are not con- are plotted in Fig.9. Solid lines stand for the city cen- sidered (NOURBAN experiment, blue line), the hourly tem- tre model (NOURBAN-blue, SLUCM-pink) and station (or- perature profile is almost the same for the city centre and for ange) data. The dashed and dash-dotted lines correspond to the two stations in the city vicinity as no urban heat island ef- the two stations in Prague’s vicinity. As seen already for the fect is modelled. In the measured data, a clear indication for monthly data (Table3), there is a negative model bias present the urban cool island effect is identifiable (with the vicinity of throughout the day. When not considering urban surfaces Prague being warmer than the centre by around 0.3 K), while (i.e. the NOURBAN experiment), this bias reaches, during very weak UCI is present in the model data (up to −0.05 K). late evening hours −3 K in city centre and around −2 K for the stations in the city’s vicinity. When the SLUCM surface 3.3 Sensitivity tests model is turned on, an evident model bias reduction is seen during afternoon and evening hours in the city centre, i.e. The urban canopy model can be configured with a whole when the UHI achieves its peak. The UHI presence is fur- range of parameters describing the geometry and the ther- momechanical properties of the surfaces typical for the ur-

www.atmos-chem-phys.net/14/12393/2014/ Atmos. Chem. Phys., 14, 12393–12413, 2014

1 1 Figure 6. The impact of urbanized surfaces on summer surface specific humidity in g kg− (left) and total precipitation rate in mm day− (right) for day (above) and nighttime (bottom) conditions. Shaded areas represent statistically significant differences on the 95 % level. from the city centers) taken from the SLUCM experiment. maximum temperature decrease of -0.5 to -0.1 K reaching 510 In our simulations, the UHI develops during noon hours and at around 9-11 a.m. reaches maximum around 10 p.m. to midnight local time at It is interesting to see that the city temperatures for con- about 0.5 K on annual average. The UHI diminishes around figuration disregarding the urban surfaces (NOURBAN, blue 6 a.m. 520 line) are lower than the temperatures accounting for urban During morning hours till noon, a weak UCI (urban cool effect in the vicinity of cities during the maximum UHI de- 515 island) develops over Berlin, Budapest and Munich, with velopment (SLUCM, orange line), by up to 0.1 K on annual average. This is consistent with our spatial results, which 8 P. Huszar et al.: Urban land-surface impact on climate

1 Figure 5. Same as Fig. 3 but for the wind velocity at 10 m in m s− . 12402 P. Huszar et al.: Urban land-surface impact on climate

Figure 6. The impact of urbanized surfaces on summer surface specific humidity in gkg−1 (left) and total precipitation rate in mmday−1 1 1 Figure 6. The impact(right) of urbanized for day (above) surfaces and night-time on summer (bottom) conditions. surface Shaded specific areas represent humidity statistically in g kg significant− (left) differences and ontotal the 95 precipitation % level. rate in mm day− (right) for day (above) and nighttime (bottom) conditions. Shaded areas represent statistically significant differences on the 95 % level. ban environment (Table1). These parameters are set glob- The change of spatial distribution of selected meteorolog- ally without any spatial variation within the domain. Further- ical parameters after introducing the modifications of the ur- more, for a given grid box (2km × 2km for our subgrid treat- ban parameters is evaluated as the difference between the from the city centers)ment taken of surface from processes), the these SLUCM parameters experiment. may, in general, correspondingmaximum SENx temperature experiment and decrease SLUCM experiment. of -0.5 to -0.1 K reaching vary from street to street, from building to building. Finally, We focused on the summer night-time average change when 510 In our simulations, the UHI develops during noon hours and at around 9-11 a.m. even choosing these parameters to match the average condi- the impacts are large, and in case of the SEN3 sensitivity reaches maximum aroundtions in urban 10 canopyp.m.to for amidnight given domain local (region) time is a chal- at experimentIt is interesting (reduced albedo) to the see average that summer the city daytime temperatures for con- about 0.5 K on annuallenging average. task. The UHI diminishes around changefiguration is shown, disregarding as albedo is relevant the urbanfor the reflected surfaces so- (NOURBAN, blue Considering the above mentioned, there is a certain degree lar radiation. 6 a.m. of uncertainty with the results originating in the estimation520 line)Figureare 10 presents lower impact than on the near temperatures surface temperature. accounting for urban During morning hoursof the SLUCM till noon, parameters. a weak To evaluate UCI this (urban uncertainty, cool an Wheneffect reducing in the building vicinity height by of 50 cities %, significant during tempera- the maximum UHI de- additional set of simulations is performed for the year 2005 ture reduction occurs up to −0.4 K over most of the large ur- 515 island) develops overwith Berlin,modified urban Budapest canopy parameters: and Munich, street width with (50 % banizedvelopment areas (cities (SLUCM, such Budapest, orange Vienna, line), Prague, by Berlin, up to 0.1 K on annual reduction); building height (50 % reduction); roof albedo Munichaverage. or Warsaw), This which is consistent corresponds to with a 30 % our reduction spatial results, which (doubled value); and anthropogenic heat release (50 % reduc- of the absolute impact on temperature (Fig.3). The reduction tion) corresponding to sensitivity experiments SEN1, SEN2, of street width by 50 % usually increases summer night-time SEN3 and SEN4, respectively. The first two parameters are temperatures up to 0.1–0.2 K, as seen especially over Berlin, of key importance in describing the city’s geometry; the third Prague, Munich, Vienna. This means an intensification of parameter is important in urban mitigation strategies for re- the UHI phenomenon by about 20 %. Lower roof albedo reducing the UHI. The fourth one is marked with high uncer- flects more solar radiation and this also affects the near sur- tainty as well, as very few estimations or measurements exist face temperatures during day. Over many cities and even over on AHR. larger areas around cities, the temperature reduction reaches −0.1 to −0.2 K, which significantly reduces the absolute im-

Atmos. Chem. Phys., 14, 12393–12413, 2014 www.atmos-chem-phys.net/14/12393/2014/ P. Huszar et al.: Urban land-surface impact on climate 9 P. Huszar et al.: Urban land-surface impact on climate 12403

Figure 7. The impact of urbanized surfaces on the winter (left) and summer (right) vertical cross-section of temperature along the 50◦ N latitude for day (above) and night-time (bottom) conditions in K averaged over the years 2005–2009. The vertical axis denotes the average Figure 7. The impact of urbanized surfaces on the winter (left) and summer (right) vertical cross-section of temperature along the 50◦ N model levels heights in metres. Shaded areas represent statistically significant differences on the 95 % level. latitude for day (above) and nighttime (bottom) conditions in K averaged over years 2005–2009. The vertical axis denotes the average model levels heights in meters. Shaded areas represent statistically significant differences on the 95 % level. pact of urban surfaces on temperature seen in Fig.3. Finally, ZPBL is affected only slightly, with a decrease typically up the AHR reduction reduces also the night-time temperatures to −20 m. Table 3. The mean 2005–2009by up to −0.2 monthlyK, not only summerover cities but 2 m overtemperatures areas far from averagedFor other over meteorological selected parameters, cities in the central spatial resultsEurope of for the NOURBAN, large urbanized centres. the sensitivity tests are not shown, only the changes for five SLUCM runs and extractedThe effects from on E-OBS the PBL observational height (ZPBL) change data. within the selected cities. For completeness, we include the near surface sensitivity tests are plotted in Fig. 11. The impact on ZPBL is temperature and ZPBL as well. The results are presented for Praguespatially noisier than for the temperature. Vienna For the 50 % build- Budapestboth day and night in Table4, where Munich bold numbers mark sta- Berlin NOURBAN SLUCMing height E-OBS reduction, NOURBAN there is anSLUCM indication of E-OBS statistically NOURBAN sig- tistically SLUCM significant E-OBS differences. NOURBAN The SLUCM table indicates E-OBS that, NOURBAN as SLUCM E-OBS − June 15.8 16.4nificant ZPBL 18.0 decrease 18.6 of up to 19.620 m over 21.0cities (especially 19.6already 20.3 seen in 21.0 Fig. 10, reducing 15.7 building 16.5 height 19.0 causes tem- 15.9 16.4 18.0 − July 18.4 18.9Berlin, Vienna 20.0 and Warsaw). 21.2 For 22.1 the 50 % 22.0 street width 22.4 re- perature 22.9 decrease, 23.0 especially 18.4 during the 19.1 night (from 21.00.03 18.4 19.0 19.0 August 16.9 17.4duction, 18.0 the ZPBL change 19.7 pattern 20.6 is even 20.0 noisier, without 20.4to − 21.00.1 K), while 21.0 reduced 16.7 street width 17.4 results in 18.0 increased 16.6 17.1 17.0 clear impact (slight increase for a few cities, but decrease for night-time temperatures (0.01 to 0.1 K). The increased roof others). A more pronounced impact is modelled for the in- albedo reduces the daytime temperatures about −0.1 K over creased roof albedo, when due to enhanced reflection and to cities in central Europe. Finally, reduced AHR causes tem- lower surface temperatures the PBL stabilizes, decreasing its perature decrease during both day and night-time in cities up showed that the impactheight. ofThis urban reduction surfaces reaches −30 is m over not large limited areas, espe- to towhen−0.1 K. the The ZPBL UHI over achieves cities decreases its peak. for reduced The build- UHI presence is fur- cially around cities. Finally, when decreasing the AHR, the ing height, increased albedo and reduced AHR, but a slight 525 the air column over large cities, but is spread over remote 545 thermore indicated by lower temperatures in the city vicin- areas as well, as seen in Fig. 3. ity in both the SLUCM experiment (by almost 0.5 K) and in For more detailed analysis, hourly station data from three the measurements (up to 1 K). When urban surfaces are not www.atmos-chem-phys.net/14/12393/2014/ Atmos. Chem. Phys., 14, 12393–12413, 2014 locations in Prague are used to compare the observed hourly considered (NOURBAN experiment, blue line), the hourly temperature variations and the modeled summer UHI. One temperature profile is almost the same for the city center and

530 station lies in the inner center of Prague, where the maxi- 550 for the two stations in the city vicinity as no urban heat is- mum UHI is expected, while two are located in the vicin- land effect is modeled. In the measured data, a clear indica- ity of the city center (about 10–15 km far). The results tion for the urban cool island effect is indentiﬁable (with the are plotted in Fig. 9. Solid lines stand for the city cen- vicinity of Prague being warmer than the center by around ter model (NOURBAN-blue, SLUCM-pink) and station (or- 0.3 K) while very weak UCI is present in the model data (up 535 ange) data. The dashed and dash-dotted lines correspond to 555 to 0.05 K). the two stations in Prague’s vicinity. As seen already for the − monthly data (Tab. 3), there is a negative model bias present 3.3 Sensitivity tests throughout the day. When not considering urban surfaces (i.e. the NOURBAN experiment), this bias reaches, during The urban-canopy model can be conﬁgured with a whole 540 late evening hours, 3 K in city center and around 2 K for range of parameters describing the geometry and the thermo- the stations in the city’s− vicinity. When the SLUCM− surface mechanical properties of the surfaces typical for urban envi- model is turned on, an evident model bias reduction is seen 560 ronment (Table 1). These parameters are set globally with- during afternoon and evening hours in the city center, i.e. out any spatial variation within the domain. Furthermore, for 10 P. Huszar et al.: Urban land-surface impact on climate 12404 P. Huszar et al.: Urban land-surface impact on climate 10 P. Huszar et al.: Urban land-surface impact on climate

Figure 8. The mean 2005–2009 diurnal 2 m temperatures variation averaged over selected cities and over its vicinity in central Europe for the NOURBANFigure (blue), 8. The SLUCM mean 2005–2009 (green –diurnal city, 2 orange m temperatures – vicinity) variation runs averaged including over vertical selected cities error and bars. over its vicinity in central Europe for Figure 8. The mean 2005–2009the NOURBAN diurnal (blue), SLUCM 2 m (greentemperatures – city, orange variation – vicinity) runs averaged including vertical over error selected bars. cities and over its vicinity in central Europe for the NOURBAN (blue), SLUCM (green – city, orange – vicinity) runs including vertical error bars.

Figure 9. The mean summer diurnal 2 m temperatures variation in Prague for the NOURBAN experiment (blue), the SLUCM experiment Figure 9. The(green) mean and summer for measurements diurnal 2(orange)m temperatures from a station variation in the centre in of Prague Prague (solid for the line) NOURBAN and two stations experiment from the vicinity (blue), of the the city SLUCM (dashed experiment (green) and forlines), measurements including vertical (orange) error bars from for athe station city centre in the series. center of Prague (solid line) and two stations from the vicinity of the city (dashed lines) including vertical error bars for the city center series increase is modelled (especially during night-time) when de- The 2 m specific humidity (q2m) tends to increase when creasing street width. lowering building heights and to decrease when reducing Figure 9. The mean summer diurnal 2 m temperatures variation− in Prague for the NOURBAN experiment (blue), the SLUCM experiment a given gridboxThe ( wind2km speed2km at 10 mfor usually our subgridincreases (up treatment to 0.3 ms of1)580 streetas well, width. as Decrease very few occurs estimations over cities as or well measurements when roof exist on (green) andsurface for measurements processes),when reducing these (orange)× building parameters from height, a station may, especially in in general, during the center the vary day- of PraguealbedoAHR. (solid is increased line) (especially and two over stations Budapest). from Finally, the vicinity with of the city (dashed lines) includingfrom street vertical totime. street, error It also bars from further for building increases the cityto over center building. each city series during Finally, both even night reducedThe AHR, change q2m of shows spatial both distribution increase and of decrease, selected but meteorolog- and daytime when reducing street width. The reduced roof usually of very small magnitude. 565 choosing these parameters to match the average conditions ical parameters after introducing the modifications of the ur- albedo causes a slight decrease of wind speed which, how- The sensitivity runs show a very small effect on total pre- in urban canopyever, is for not a statistically given domain significant (region) in most is a cases. challenging The same cipitation.ban parameters When reducing is evaluated building as heights, the difference the precipita- between the task. is true for the impact of AHR reduction on the wind speed,585 tioncorresponding decreases significantly SENxexperiment only over Budapest and duringSLUCM day- experiment. a given gridbox (2km 2km for our subgrid treatment of 580 as well, as very few estimations or measurements exist on although the numbers indicate slight decrease. time (−0.4 mmday−1). Over the same city, the reduced street surface processes),Considering these the× above parameters mentioned, may, there in is general, a certain degree vary WeAHR. focused on the summer nighttime average change when of uncertainty of the results originating in the estimation of the impacts are large, and in case of the SEN3 sensitivity from street to street, from building to building. Finally, even The change of spatial distribution of selected meteorolog- 570 the SLUCM parameters. To evaluate this uncertainty, an ad- experiment (reduced albedo) the average summer daytime Atmos. Chem. Phys., 14, 12393–12413, 2014 www.atmos-chem-phys.net/14/12393/2014/ 565 choosingditional these parameters set of simulations to match is performed the average for year conditions 2005 with changeical parameters is shown, as after albedo introducing is relevant the for themodifications reflected so- of the ur- in urban canopymodified for urban a given canopy domain parameters: (region) street widthis a challenging (50 % reduc- 590 larban radiation. parameters is evaluated as the difference between the task. tion), building height (50 % reduction), roof albedo (doubled585 correspondingFigure 10 presents SENx impact experiment on near surface and SLUCM temperature. experiment. Consideringvalue) the and above anthropogenic mentioned, heat release there (50is a % certain reduction) degree cor- WhenWe focused reducing on building the summer height by nighttime 50 %, significant average tempera- change when of uncertainty575 responding of the to results sensitivity originating experiments in SEN1, the estimation SEN2, SEN3 of turethe reduction impacts occurs are large, up to and0.4 K inover case most of of the the SEN3 large ur- sensitivity and SEN4, respectively. The first two parameters are of key banized areas (cities such Budapest,− Vienna, Prague, Berlin, 570 the SLUCM parameters. To evaluate this uncertainty, an ad- experiment (reduced albedo) the average summer daytime importance in describing the city’s geometry, the third pa- 595 Munich or Warsaw), which corresponds to a 30 % reduction ditional setrameter of simulations is important in is urban performed mitigation for strategies year 2005 for reduc- with ofchange the absolute is shown, impact as on temperaturealbedo is relevant (Fig. 3). The for reduction the reflected so- modified urbaning the canopy UHI. The parameters: fourth one is street marked width with high (50 uncertainty% reduc- 590 oflar street radiation. width by 50 % usually increases summer nighttime tion), building height (50 % reduction), roof albedo (doubled temperaturesFigure 10 up presentsto 0.1–0.2 K impact, as seen on especially near surface over Berlin, temperature. value) and anthropogenic heat release (50 % reduction) cor- When reducing building height by 50 %, significant tempera- 575 responding to sensitivity experiments SEN1, SEN2, SEN3 ture reduction occurs up to 0.4 K over most of the large ur- and SEN4, respectively. The first two parameters are of key banized areas (cities such Budapest,− Vienna, Prague, Berlin, importance in describing the city’s geometry, the third pa- 595 Munich or Warsaw), which corresponds to a 30 % reduction rameter is important in urban mitigation strategies for reduc- of the absolute impact on temperature (Fig. 3). The reduction ing the UHI. The fourth one is marked with high uncertainty of street width by 50 % usually increases summer nighttime temperatures up to 0.1–0.2 K, as seen especially over Berlin, P. Huszar et al.: Urban land-surface impact on climate 11

P. Huszar et al.: Urban land-surface impact on climate 12405

Figure 10. Sensitivity test: the impact of 50 % reduction of building size (upper left), 50 % reduction of street width (upper right), two times Figure 10. Sensitivityhigher test: roof the albedo impact (bottom of left) 50 and % 50 reduction % reduction of AHR building (bottom right)size on (upper the 2005 left), summer 50 average % reduction near surface temperatureof street for width night- (upper right), 2 times higher roof albedo (bottomtime conditions left) and (SEN1, 50 SEN2 % reduction and SEN4) and of daytime AHR conditions (bottom (SEN3) right) in onK. Shaded 2005 areas summer represent average statistically near significant surface differences temperature for nighttime conditions (SEN1, SEN2on the and 95 % SEN4) level. and daytime conditions (SEN3) in K. Shaded areas represent statistically significant differences on the 95 % level. width has significant impact on precipitation, with decrease Figure 13 presents the impact of urban surfaces corre- during daytime but increase during the night. Reducing roof sponding only to Prague on the near surface temperature (in- albedo leads to significant decrease of precipitation only over cluding the SUBBATS treatment). The statistically signifi- Prague and Budapest, over other cities the changes are am- cant temperature impact is limited to a small region (up to biguous. Finally, reduced AHR leads to a statistically signif- 150 km in diameter) around Prague for both day and night. icant decrease of daytime precipitation only over Vienna and The maximum impact corresponds to the Prague city cen- Berlin (up to −1 mmday−1). tre and reaches 0.4 and 1.0 K for day and night, respectively, It is also of interest how the results are influenced by which is, for night-time, slightly less compared to the case the application of the subgrid surface treatment using the when all the urban surfaces are considered (Fig.3). RegCM4.2 SUBBATS feature (at 2km × 2km resolution in our case). Figure 12 shows the mean 2005–2009 summer temperature change for day and night evaluated without us- 4 Discussion and conclusions ing SUBBATS. The daytime impact on near surface temper- The regional impact of urban surfaces on the meteorologi- ature is relatively smooth and large urban centres are dif- cal conditions over central Europe was evaluated, using the ficult to identify. Maximum impact exceeds 0.2 K, but can regional climate model RegCM4.2 extended with a single- reach 1 K in Italy. During night, the urban centres become layer urban canopy model. We focused on the long-term ef- well visible with usually more than 1 K impact (often more fects performing 5 year simulations. than 1.5 K, i.e. the impact is large over grid boxes with urban In terms of temperature, the largest impacts are modelled land use category (Fig.2, left), but is significant even over during summer night-time with up to 1.5 K higher tempera- grid boxes without urban land use category, often exceeding tures than without considering urban surfaces. This is con- 0.1 K). sistent with the maximum UHI development during evening

www.atmos-chem-phys.net/14/12393/2014/ Atmos. Chem. Phys., 14, 12393–12413, 2014

Figure 11. Same as Fig. 10 but for the PBL height in m.

Prague, Munich, Vienna. This means an intensiﬁcation of 605 pact of urban surfaces on temperature seen in Fig. 3. Finally,

600 the UHI phenomenon by about 20 %. Lower roof albedo re- the AHR reduction reduces also the nighttime temperatures ﬂects more solar radiation and this also affects the near sur- by up to 0.2 K not only over cities but over areas far from face temperatures during day. Over many cities and even over large urbanized− centers. larger areas around cities, the temperature reduction reaches The effects on the PBL height (ZPBL) change within the 0.1 to 0.2 K, which signiﬁcantly reduces the absolute im- 610 sensitivity tests are plotted in Fig. 11. The impact on ZPBL is − − P. Huszar et al.: Urban land-surface impact on climate 11

12 P. Huszar et al.: Urban land-surface impact on climate

Table 4. Sensitivity tests: changes in selected meteorological parameters (T2 m – 2 m temperature [K], Zpbl – planetary boundary layer height Figure 10. Sensitivity test: the impact1 of 50 % reduction of building size (upper left),1 50 % reduction of street width (upper right),1 2 times [m], V10 m – wind speed at 10 m s− [m], Q2 m – specific humidity at 2 m [g kg− ], Ptot – total precipitation rate [mm day− ]) over five citieshigher due roof to albedo changes (bottom in urban left) canopy and 50 %parameters reduction of of SLUCM AHR (bottom for day right) and on nighttime 2005 summer in summer average 2005. near Bold surface numbers temperature indicate for statistically nighttime significantconditions differences(SEN1, SEN2 on andthe 95 SEN4) % level. and daytime conditions (SEN3) in K. Shaded areas represent statistically significant differences on the 95 % level. 12406Prague Vienna P. Budapest Huszar et al.: Urban land-surface Munich impact on Berlin climate JJA JJA JJA JJA JJA JJA JJA JJA JJA JJA day night day night day night day night day night

SEN1 T2 m 0.02 0.03 0.00 0.06 0.00 0.13 0.06 0.04 0.05 0.28 − − − − − − − − b.height Zpbl 4.52 1.01 8.83 1.49 4.12 1.34 13.5 0.9 9.57 18.9 − − − − − − − − − 50 % V10 m 0.06 0.03 0.11 0.05 0.09 0.02 0.01 0.01 0.31 0.06 − − Q2 m 0.02 0.02 0.01 0.00 0.00 0.01 0.05 0.00 0.04 0.08 − − − Ptot 0.39 0.44 0.23 0.27 0.40 0.10 0.88 0.12 0.10 0.23 − − − − SEN2 T2 m 0.01 0.03 0.02 0.02 0.02 0.01 0.00 0.03 0.08 0.10 − str.width Zpbl 0.40 4.35 13.6 3.47 4.13 4.67 4.93 0.84 7.55 8.73 − − 50 % V10 m 0.06 0.09 0.10 0.14 0.12 0.10 0.05 0.02 0.30 0.31 Q2 m 0.01 0.00 0.01 0.02 0.00 0.00 0.01 0.00 0.05 0.00 − − − − − − Ptot 0.46 0.60 0.11 0.03 0.41 0.29 0.70 0.39 0.05 0.07 − − − − − − − SEN3 T2 m 0.11 0.20 0.11 0.02 0.06 0.01 0.08 0.01 0.13 0.00 − − − − − − − − 2 x Zpbl 17.5 1.11 31.8 4.86 11.0 8.46 28.7 3.00 24.8 1.92 − − − − − − roof alb. V10 m 0.02 0.03 0.02 0.02 0.02 0.01 0.04 0.01 0.05 0.00 − − − − − − − − − Q2 m 0.02 0.02 0.01 0.01 0.05 0.06 0.03 0.00 0.00 0.04 − − − − − − Ptot 0.17 1.00 0.10 0.02 0.57 0.36 0.56 0.33 0.05 0.09 − − − − − SEN4 T2 m 0.07 0.01 0.05 0.03 0.04 0.05 0.1 0.04 0.07 0.10 − − − − − − − − − AHR Zpbl 15.6 5.38 8.71 5.63 6.60 8.57 37.7 3.15 17.6 9.14 − − − − − 50 % V10 m 0.02 0.01 0.02 0.01 0.03 0.02 0.03 0.02 0.01 0.05 − − − − − − Q2 m 0.04 0.02 0.00 0.00 0.00 0.02 0.00 0.02 0.02 0.09 − − − − − Ptot 0.45 0.37 1.00 0.17 0.19 0.18 0.67 0.02 0.59 0.19 − − − − − −

Figure 11. Same as Fig. 10 but for the PBL height in m. Figure 11. Same as Fig. 10 but for the PBL height in m.

Prague, Munich, Vienna. This means an intensiﬁcation of 605 pact of urban surfaces on temperature seen in Fig. 3. Finally,

Figure 12. The impact of urbanized surfaces on the summer day (left) and night (right) near surface temperature in K averaged over years Figure 12. The impact2005–2009 of urbanized without SUBBATS, surfaces i.e. on at 10 the km summer× 10 km model day resolution. (left) and Shaded night areas represent (right) statistically near surface significant temperature differences on thein K averaged over years 2005–2009 without SUBBATS,95 % level. i.e. at 10km 10km model resolution. Shaded areas represent statistically significant differences on the 95 % level. × and night hours seen in previous studies for European cities son et al., 2008). In winter, the impact is significant mainly (Eliasson and Holmer, 1990 for Gothenburg; Pichierri et al., during daytime (seen similarly in Struzewska and Kaminski 2012 for Milan; Giannaros and Melas, 2012 for Thessaloniki; (2012)), which is probably due to the governing role of AHR spatially noisier thanGiannaros for the et temperature. al., 2013 for Athens) For or forthe US 50 cities % (e.g. build- Ole- beingZPBL stronger is during affected the day. only Indeed, slightly, the hourly with profiles a in decrease typically up ing height reduction, there is an indication of statistically to 20 m. − significant ZPBL decreaseAtmos. Chem. of up Phys., to 14,20 12393–m12413over, 2014 cities (espe- For other www.atmos-chem-phys.net/14/12393/2014/ meteorological parameters, the spatial results of − cially Berlin, Vienna and Warsaw). For the 50 % street width 625 the sensitivity tests are not shown, only the changes for five

615 reduction, the ZPBL change pattern is even noisier, without selected cities. For completeness, we include the near sur- clear impact (slight increase for a few cities, but decrease for face temperature and ZPBL as well. The results are presented others). A more pronounced impact is modeled for the infor both day and night in Table 4, where bold numbers mark creased roof albedo, when due to enhanced reﬂection and to statistically signiﬁcant differences. The table indicates that, lower surface temperatures the PBL stabilizes, decreasing its 630 as already seen in Fig. 10, reducing building height causes 620 height. This reduction reaches 30 m over large areas, espe- temperature decrease, especially during night (from 0.03 cially around cities. Finally, when− decreasing the AHR, the to 0.1 K) while reduced street width results in increased− nighttime− temperatures (0.01 to 0.1 K). The increased roof P. Huszar et al.: Urban land-surface impact on climate 12407

Table 4. Sensitivity tests: changes in selected meteorological parameters (T2 m – 2 m temperature [K], Zpbl – planetary boundary layer height −1 −1 −1 [m], V10 m – wind speed at 10 ms [m], Q2 m – specific humidity at 2 m [gkg ], Ptot – total precipitation rate [mmday ]) over five cities due to changes in urban canopy parameters of SLUCM for day and night-time in summer 2005. Bold numbers indicate statistically significant differences on the 95 % level.

Prague Vienna Budapest Munich Berlin JJA JJA JJA JJA JJA JJA JJA JJA JJA JJA day night day night day night day night day night

SEN1 T2 m −0.02 −0.03 −0.00 −0.06 0.00 −0.13 −0.06 −0.04 0.05 −0.28 b.height Zpbl −4.52 −1.01 −8.83 −1.49 −4.12 −1.34 −13.5 0.9 −9.57 −18.9 50 % V10 m 0.06 −0.03 0.11 0.05 0.09 −0.02 0.01 0.01 0.31 0.06 Q2 m 0.02 0.02 0.01 −0.00 −0.00 −0.01 0.05 0.00 0.04 0.08 Ptot 0.39 −0.44 −0.23 0.27 −0.40 0.10 0.88 −0.12 0.10 0.23

SEN2 T2 m 0.01 0.03 0.02 0.02 −0.02 0.01 0.00 0.03 0.08 0.10 str.width Zpbl −0.40 4.35 −13.6 3.47 4.13 4.67 4.93 0.84 7.55 8.73 50 % V10 m 0.06 0.09 0.10 0.14 0.12 0.10 0.05 0.02 0.30 0.31 Q2 m −0.01 −0.00 0.01 −0.02 0.00 −0.00 0.01 0.00 −0.05 −0.00 Ptot −0.46 −0.60 0.11 −0.03 −0.41 0.29 0.70 −0.39 −0.05 −0.07

SEN3 T2 m −0.11 −0.20 −0.11 0.02 −0.06 0.01 −0.08 −0.01 −0.13 −0.00 2 x Zpbl −17.5 1.11 −31.8 4.86 −11.0 8.46 −28.7 3.00 −24.8 −1.92 roof alb. V10 m −0.02 −0.03 −0.02 0.02 −0.02 −0.01 −0.04 −0.01 −0.05 −0.00 Q2 m 0.02 0.02 −0.01 −0.01 −0.05 −0.06 0.03 −0.00 −0.00 0.04 Ptot 0.17 −1.00 0.10 0.02 −0.57 0.36 0.56 −0.33 −0.05 −0.09

SEN4 T2 m −0.07 −0.01 −0.05 0.03 −0.04 −0.05 −0.1 −0.04 −0.07 −0.10 AHR Zpbl −15.6 5.38 8.71 5.63 −6.60 8.57 −37.7 3.15 −17.6 −9.14 P. Huszar et al.: Urban land-surface50 % V impact10 m −0.02 on 0.01 climate−0.02 0.01 0.03 −0.02 −0.03 −0.02 0.01 −0.05 13 Q2 m 0.04 0.02 −0.00 −0.00 −0.00 −0.02 0.00 −0.02 0.02 0.09 Ptot 0.45 −0.37 −1.00 −0.17 −0.19 0.18 0.67 −0.02 −0.59 0.19

Figure 13. Impact of urbanized surfaces corresponding to Prague on the near surface temperature including the SUBBATS surface treatment Figure 13. Impact of urbanized(2km × 2km resolution). surfaces Shaded corresponding areas represent to statistically Prague significant on the differences near surface on the 95 temperature % level. including the SUBBATS surface treatment (2km 2km resolution). Shaded areas represent statistically significant differences on the 95 % level. × our simulations assume that significantly larger heat is re- temperature during winter (seen especially for Berlin, Prague leased from vehicular traffic during the day than during the or Budapest) is probably the result of stabilized PBL and re- albedo reduces the daytimenight; and that temperatures AHR from indoor about heating peaks0.1 duringK over morn- ducedficult wind to speeds identify. (see further), Maximum which can lead impact to develop- exceeds 0.2 K, but can ing hours as the heat from nocturnal heating− is released with ment of stronger winter inversion layers. 635 cities in central Europe.a lag due Finally, to thermal reduced conduction through AHR the causes roofs and tem- walls reachIt is also 1 foundK in that Italy. enhanced During vertical mixing night, in summerthe urban centres become (Sailor and Lu, 2004). As a result, daytime AHR is slightly causes the impact of urban surfaces on temperature to spread perature decrease during both day and nighttime in cities up 675 well visible with usually more than 1 impact (often more stronger than during the night. Feng et al.(2012) also found across the whole PBL. The modelled decrease of temperatureK to 0.1 K. The ZPBLAHR over to be cities the most decreases important factor for influencingreduced the build- urban overthan the PBL 1.5 canK, be i.e. explained the impact by the increased is large vertical over lapse gridboxes with urban ing− height, increasedimpact albedo on winter and temperatures. reduced The AHR, night-time but decrease a slight of landuse category (Fig. 2, left), but is significant even over increase is modeled (especially during nighttime) when de- gridboxes without urban landuse category, often exceeding 640 creasing street width.www.atmos-chem-phys.net/14/12393/2014/0.1 K). Atmos. Chem. Phys., 14, 12393–12413, 2014 1 The wind speed at 10 m usually increases (up to 0.3 m s− ) 680 Figure 13 presents the impact of urban surfaces corre- when reducing building height, especially during daytime. It sponding only to Prague on the near surface temperature also further increases over each city during both night and (incl. the SUBBATS treatment). The statistically significant daytime when reducing street width. The reduced roof albedo temperature impact is limited to a small region (up to 150 km 645 causes a slight decrease of wind speed which, however, is not in diameter) around Prague for both day and night. The statistically significant in most cases. The same is true for the 685 maximum impact corresponds to the Prague city centre and impact of AHR reduction on the wind speed, although the reaches 0.4 and 1.0 K for day and night, respectively, which numbers indicate slight decrease. is, for nighttime, slightly less compared to the case when all The 2 m specific humidity (q2m) tends to increase when the urban surfaces are considered (Fig. 3). 650 lowering building heights and to decrease when reducing street width. Decrease occurs over cities as well when roof albedo is increased (especially over Budapest). Finally, with 4 Discussion and conclusions reduced AHR, q2m shows both increase and decrease, but usually of very small magnitude. 690 The regional impact of urban surfaces on the meteorological conditions over central Europe was evaluated, using the 655 The sensitivity runs show a very small effect on total precipitation. When reducing building heights, the precipitation regional climate model RegCM4.2 extended with a single- decreases significantly only over Budapest during daytime layer urban canopy model. We focused on the long term ef- 1 fects performing 5 year simulations. ( 0.4 mm day− ). Over the same city, the reduced street width− has significant impact on precipitation, with decrease 695 In terms of temperature, the largest impacts are modeled during summer nighttime with up to 1.5 K higher tempera- 660 during daytime but increase during night. Reducing roof albedo leads to significant decrease of precipitation only over tures than without considering urban surfaces. This is con- Prague and Budapest, over other cities the changes are am- sistent with the maximum UHI development during evening biguous. Finally, reduced AHR leads to a statistically signif- and night hours seen in previous studies for European cities icant decrease of daytime precipitation only over Vienna and 700 (Eliasson and Holmer, 1990 for Goteborg; Pichierri et al., 1 2012 for Milan; Giannaros and Melas, 2012 for Thessaloniki; 665 Berlin (up to 1 mm day− ). It is also of− interest how the results are influenced by Giannaros et al., 2013 for Athens) or for US cities (e.g. Ole- the application of the subgrid surface treatment using the son et al., 2008). In winter, the impact is significant mainly during daytime (seen similarly in Struzewska and Kaminski RegCM4.2 SUBBATS feature (at 2km 2km resolution in our case). Figure 12 shows the mean× 2005–2009 summer 705 (2012)), which is probably due to governing role of AHR being stronger during the day. Indeed, the hourly profiles in our 670 temperature change for day and night evaluated without using SUBBATS. The daytime impact on near surface temper- simulations assume that significantly larger heat is released ature is relatively smooth and large urban centres are dif- from vehicular traffic during the day than during night and that AHR from indoor heating peaks during morning hours 14 P. Huszar et al.: Urban land-surface impact on climate

12408 P. Huszar et al.: Urban land-surface impact on climate et al. (2013) attributed this wind stilling to increased surface

745 roughness. A similar conclusion was made by Vautard et al. (2010) analyzing the observed Northern Hemispheric wind stilling. Another factor inﬂuencing the urbanization impact on wind is the destabilization of urban boundary layer (UBL) due to higher urban temperatures (seen also in connection

750 with the PBL increase) which on the other hand leads to enhanced winds. These competing effects may result in slight wind speed decrease over cities and a small wind speed increase around these areas during daytime. The increase in wind speed during nighttime can be re-

755 lated to decreased nocturnal stability due to the presence of UHI. Another contributing factor could be the generation of urban-breeze circulation, i.e. formation of convergent motions towards the city as a result of thermally induced hor- Figure 14. The cool island effect: the impact of urbanized surfaces on the summer morning 8–11 a.m. near surface temperature in K averaged Figureover years 2005–2009. 14. The Shaded areas cool represent island statistically effect: signiﬁcant differences the impact on the 95 % level. of urbanized surfaces izontal pressure gradient (Hidalgo et al., 2010), however, it

onrate the due to summer higher PBL temperature, morning resulting 8-11 in enhanced a.m.roughness. near A surface similar conclusion temperature was made by Vautard in et al. K 760 is probably not resolvable with our dynamical 10 km x 10 convective motion, as discussed by Collier(2006). (2010) analysing the observed Northern Hemispheric wind averagedAccording to over our results, years the diurnal 2005–2009. temperature range re- Shadedstilling. Another areas factor represent influencing the urbanization statistically impact km resolution. Indeed, most of the studies dealing with this duction due to urbanized surfaces is about −0.8 K in summer on wind is the destabilization of urban boundary layer (UBL) significantand −0.2 K in winter, differences which is lower than on values the found 95 in %due level. to higher urban temperatures (seen also in connection type of circulation used much finer horizontal step, e.g. the Trusilova et al.(2008)( −1.2 and −0.7 K, respectively). with the PBL increase), which on the other hand leads to en- In our simulations, the PBL height increases in both winter hanced winds. These competing effects may result in slight mentioned study calculated on a 500 m x 500 m grid. and summer daytime, and summer night-time, but decrease wind speed decrease over cities and a small wind speed in- is modelled over cities in winter night-time. The daytime crease around these areas during daytime. Enhanced roughness in urban environment may play a increase is explained by enhanced turbulent mixing due to The increase in wind speed during night-time can be re- the urban morphology as already found by Martilli(2002); lated to decreased nocturnal stability due to the presence of 710 765 asAngevine the et heat al.(2003); from Rotach et al.nocturnal(2005); Collier(2006 heating). UHI. Another is released contributing factor with could be athe generationlag due of governing factor in the winter nigh-time decrease of winds, The summer night-time PBL increase is limited over cities urban-breeze circulation, i.e. formation of convergent mo- toand thermal can be the result of conduction warmer air in the nocturnal through PBL due tions the towards roofs the city as and a result of walls thermally induced (Sailor hor- resulting in lower modeled PBL heights, which was also doc- to UHI. In winter night-time, a slight PBL height reduction is izontal pressure gradient (Hidalgo et al., 2010), however, it andmodelled. Lu, This 2004). is attributable to As decreased a result, wind speeds (seedaytimeis probably AHR not resolvable is withslightly our dynamical stronger 10 km x 10 umented by Hou et al. (2013). A different summer nocturnal below), which leads to decreased mixing. km resolution. Indeed, most of the studies dealing with this thanThe wind during speed changes the in both night. winter and summer Feng day- ettype al. of circulation (2012) used much also finer horizontal found step, AHR e.g. the behavior is simulated for northern Italy (urban areas within time are characterized with a slight decrease over cities. mentioned study calculated on a 500 m × 500 m grid. toSimilar be decreases the most were found important by Klaic´ et al.(2002 factor). Hou influencingEnhanced roughness in the urban urbanenvironment impactmay play a the Po valley). Here, even in summer nighttime, wind speed et al.(2013) attributed this wind stilling to increased surface governing factor in the winter night-time decrease of winds, 715 on winter temperatures. The nighttime decrease of tempera- 770 decrease is modelled, again, probably governed by increased tureAtmos. during Chem. Phys., winter14, 12393–12413 (seen, 2014 especially for www.atmos-chem-phys.net/14/12393/2014/ Berlin, Prague or Bu- surface drag. A similar result was recently found for another dapest) is probably the result of stabilized PBL and reduced mediterranean city, Lisbon (Portugal) by Lopes et al. (2012). wind speeds (see further) which can lead to development of Apart from no significant changes in winter precipitation stronger winter inversion layers. (which we do not present in this study), our results show very

720 It is also found that enhanced vertical mixing in summer 775 small impact on also the night summer precipitation. How- causes the impact of urban surfaces on temperature to spread ever, a statistically signiﬁcant decrease of rainfall is modeled across the whole PBL. The modeled decrease of temperature in summer during day. Compared to Trusilova et al. (2008), over the PBL can be explained by the increased vertical lapse our results indicate a bit stronger precipitation reduction: up rate due to higher PBL temperature, resulting in enhanced to 20–30 % for summer precipitation (even stronger, 50 % re-

725 convective motion, as discussed by Collier (2006). 780 duction in summer daytime) compared to about 20 % reduc- According to our results, the diurnal temperature range re- tion found in the mentioned study. This decrease as mostly duction due to urbanized surfaces is about -0.8 K in summer driven by the convective precipitation - the large scale pre- and -0.2 K in winter, which is lower than values found in cipitation was not perturbed statistically significantly in our Trusilova et al. (2008) (-1.2 and -0.7 K, respectively). simulations. There has been evidence in other studies (e.g. 730 In our simulations, the PBL height increases in both win- 785 Kaufmann et al., 2007) that within urban environment, the ter and summer daytime, and summer nighttime, but decrease rain formation is suppressed, due to decreased availability is modeled over cities in winter nighttime. The daytime in- of moisture. Indeed, our results suggest significant reduc- crease is explained by enhanced turbulent mixing due to tion of specific humidity in the whole PBL over large ar- the urban morphology as already found by Martilli (2002); eas not limited only to urban centers. The Grell convective

735 Angevine et al. (2003); Rotach et al. (2005); Collier (2006). 790 scheme (Grell, 1993) with the Fritsch and Chappell closure The summer nighttime PBL increase is limited over cities (Fritsch and Chappell, 1980) used in this study considers that and can be the result of warmer air in the nocturnal PBL due the precipitation rate is proportional to the updraft mass ﬂux to UHI. In winter nighttime, a slight PBL height reduction is which is further proportional to the buoyant energy avail- modeled. This is attributable to decreased wind speeds (see able for convection (ABE). With higher lapse rate the ABE

740 below), which leads to decreased mixing. 795 increases but the lower humidity in the PBL is counteract- The wind speed changes in both winter and summer day- ing and suppresses ABE. In our simulations, this latter ef- time are characterized with a slight decrease over cities. fect dominates. In summary, even with potentially stronger Similar decreases were found by Klaic´ et al. (2002). Hou vertical convective motion caused by increased temperature P. Huszar et al.: Urban land-surface impact on climate 12409 resulting in lower modelled PBL heights, which was also ness in our simulations. However, after applying the SLUCM documented by Hou et al.(2013). A different summer noc- urban canopy treatment, the negative bias in summer is re- turnal behaviour is simulated for northern Italy (urban areas duced or, in some cases, vanishes entirely. Improvement in within the Po valley). Here, even in summer night-time, wind model performance is evident from the comparison of di- speed decrease is modelled, again, probably governed by in- urnal temperature variation with observed temperatures as creased surface drag. A similar result was recently found for well. The negative model bias is significantly reduced during another mediterranean city, Lisbon (Portugal) by Lopes et al. afternoon and evening hours when the urban meteorological (2012). effects are most prominent. Apart from no significant changes in winter precipitation It is important to emphasize that, regarding the tempera- (which we do not present in this study), our results show very ture, the impact of urban surfaces is not identical to the UHI small impact on also the night summer precipitation. How- magnitude, as the UHI is defined as the temperature differ- ever, a statistically significant decrease of rainfall is mod- ence between the city centre and its urban effect-free vicin- elled in summer during daytime. Compared to Trusilova et ity. We demonstrated in our simulations that the cities im- al.(2008), our results indicate a slightly stronger precipita- pact temperatures over rural areas as well. Hence, the ref- tion reduction: up to 20–30 % for summer precipitation (even erence state corresponding to the absence of urban surfaces stronger, 50 % reduction in summer daytime) compared to (the NOURBAN experiment) is colder than the reference for about 20 % reduction found in the mentioned study. This de- the UHI magnitude calculation at a particular city. This was crease was mostly driven by the convective precipitation - well seen in the plots showing the hourly temperature varia- the large scale precipitation was not perturbed statistically tion over selected cities where the temperatures in the city’s significantly in our simulations. There has been evidence in vicinity are higher than if there were no cities at all. Then, the other studies (e.g. Kaufmann et al., 2007) that within ur- annual mean UHI magnitude is 0.5–1 K. Unger et al.(2011) ban environment, the rain formation is suppressed, due to provide information on the annual mean UHI from a mid- decreased availability of moisture. Indeed, our results sug- dle sized central European city of Novi Sad in Serbia and gest significant reduction of specific humidity in the whole they find UHI up to 4 K. However, this value is representa- PBL over large areas not limited only to urban centres. The tive of the very centre of the city of a few square kilometres, Grell convective scheme (Grell, 1993) with the Fritsch and while in our simulations we averaged the temperature from Chappell closure (Fritsch and Chappell, 1980) used in this a 10km × 10km grid box. If the UHI values from the afore- study considers that the precipitation rate is proportional to mentioned study are averaged over such an area around the the updraft mass flux, which is further proportional to the city centre, the values decrease to 1–2 K, which is closer to buoyant energy available for convection (ABE). With higher our results. The same is true for Bottyan and Unger(2003) lapse rate the ABE increases but the lower humidity in the who provide mean maximum UHI intensity of 2.1 and 3.1 K PBL is counteracting and suppresses ABE. In our simula- for the Hungarian city of Szeged during heating and non- tions, this latter effect dominates. In summary, even with po- heating season, respectively. Based on satellite data, Pon- tentially stronger vertical convective motion caused by in- grácz et al.(2010) found the 2001–2003 mean night-time creased temperature lapse rate, less available moisture leads UHI values in central European cities to be around 2 K. They to reduced precipitation formation in the end. Similar results defined the city as a 15 km radius circle around the centre and conclusions are drawn in Trusilova et al.(2008) for their and a 15–25 km belt around represented the rural areas for CENTER_EU region, which largely overlaps with our do- determining the UHI. This corresponds well to our definition main. and suggests an underestimation of UHI in our simulations. Enhancement of precipitation downwind of urban areas A more pronounced UHI was also simulated by Fallmann was documented by several measurement or model studies et al.(2013) for Stuttgart (around 2.5 K, compared to our (Shepherd et al., 2002; Jauregui, 1991; Changon et al., 1991; value of about 0.7 K). They, however, used a much finer res- Rozoff et al., 2003). This enhancement is seen in our results olution of 1km × 1km and the city averaged UHI intensity only around a few cities, but in most of the cases they are (1.3 K), which is a more comparable quantity with our val- not statistically significant. The comparison is however prob- ues, is again closer to our result. It should be noted that their lematic, as already pointed out by Trusilova et al.(2008): the experiments were carried out for an exceptionally warm pe- mentioned studies focused on tropical climate and further- riod from 11 to 18 August during the 2003 heat wave, while more, they compared the downwind with the upwind precipi- we averaged over summers from 5 years. tation of the city, while we compared the rainfall correspond- The measured data over Prague revealed the UCI effect ing to the present state of urbanization with a hypothetical as well, which was also reproduced by the model, but to a state with no urban coverage, i.e. two different states of at- much lesser extent. Fig. 14, illustrating the spatial distribu- mospheric circulation. tion of the morning temperature response to urban coverage, The comparison of monthly temperature data over cities shows an overall warming already seen during both night with measurements reveals a negative model bias in summer and daytime conditions but also indicates a cooling for larger months, which is probably due to overestimation of cloudi- cities, especially over the western part of the domain. The re- www.atmos-chem-phys.net/14/12393/2014/ Atmos. Chem. Phys., 14, 12393–12413, 2014 12410 P. Huszar et al.: Urban land-surface impact on climate sults suggest presence of the UCI effect also for other cities. that can reach 0.2 K, and which could be attributed to the However, the cooling is usually eliminated by counteracting remote impact of UHI. However, Fig.3 shows that for night- warming seen all over the domain. Therefore the results are time conditions over areas with minor urbanization, such im- either insignificant (Munich, Budapest) or the they show only pact is slightly higher (by 0.05 K). This indicates that these a reduction of temperature impact (Brno – Czech Republic, small urban elements (smaller cities, suburban and partly ur- Poznan – Poland). banized areas) can contribute to the impact. This conclusion The sensitivity analysis for the summer meteorological could partially affect, at least in the areas with highly popu- conditions showed a rather limited dependence on the model lated urbanized areas as in Europe, the temperature increase setup of the geometric and radiative characteristics of the under global warming, supposing the rapid development of urban environment. Specifically, the temperature responses the urbanization in the region. to significant changes of geometry parameters such as street The second question relates to the remoteness of the ur- width or building height are of the order 0.01 K (up to 0.1 K ban impact, i.e. how purely rural areas can be influenced by in a few cases). The same is true for the AHR. Somewhat nearby (or even more distant) cities. In Fig. 13, the impact of stronger temperature response was detected for the build- Prague is apparently limited to the surrounding region of no ing roof albedo value. The changes of the boundary layer more than 150 km diameter, elsewhere the impact is statisti- height can be as high as a few tens of metres when perturb- cally insignificant (except a little noise which has not been ing the key SLUCM parameters in our simulations, which eliminated using 5 years long simulations and 95 % signif- is closer to the average modelled ZPBL change (Fig.4) and icance threshold). In conclusion, it can be summarized that reveals relatively large sensitivity of ZPBL to urban param- over remote areas the modelled temperature impact is a com- eters. Similarly, a substantial dependence of the 10 m wind bination of the distant impact of surrounding large urbanized velocity on the urban geometry parameters is modelled (with areas and the local impact of the few small urbanized sub- changes in the order of 0.1 ms−1), which is comparable to grid-boxes located in the corresponding grid box. The remote the absolute impact of urban surfaces. The roof albedo and impact is probably caused by the advection of warm air orig- the AHR had a very little impact on the wind speed. The de- inating from UHI over non-urbanized areas. This is well ob- pendence of the impact on the humidity is small, but often served especially during daytime (in summer), when wind statistically significant, especially when reducing the AHR. speed is higher, resulting in smooth, almost uniform urban Finally, precipitation can be impacted largely when changing impact on temperature across the domain. Feng et al.(2012) the SLUCM parameters (up to −1 mmday−1 when reducing simulated the long-term regional impact of urbanization and AHR), but these changes are usually insignificant in a sta- AHR and found values similar to our study: over most of the tistical sense. Therefore, it can be concluded that the results eastern China exceeding 1 K for the summer average. concerning the impact of urbanized surfaces on the long-term The question remains what improvement of the results meteorological conditions over central Europe are, with re- can be achieved by using a more sophisticated urban canopy spect to the overall uncertainty of parameters settings, quite model, such as the multilayer BEP (Building Energy Param- robust. Efforts aiming to set the urban parameters more pre- eterization) model (Martilli et al., 2002). It will certainly bet- cisely with the eventuality of using 2-D distribution of these ter describe the 3-D nature of the heat, moisture and momen- parameters will change the overall picture probably only a lit- tum exchange between buildings and street-canyon, impact- tle. ing the thermodynamic structure of the urban roughness sub- Another important conclusion regarding the impact of ur- layer and hence the lower part of the urban boundary layer. ban surfaces on temperature is that in central Europe it is not Although a multi-layer urban canopy model represents an in- limited to large cities only, but can be significant over remote crease of computational demand compared to SLUCM, fu- areas with rather minor urban development as well. A ques- ture work should focus in this direction and apply such mod- tion arises whether this impact is the result of the presence of els in long-term climate simulations in combination with re- smaller cities, i.e. a few urbanized subgrid boxes in the cor- gional climate models, in order to obtain a more accurate pic- responding 10km × 10km grid boxes, or whether urban sur- ture about the climatic impact of urbanization on a regional faces have distant influence on non-urbanized areas far from scale. them. Figure2 (right) shows that at 2km × 2km resolution, The results obtained in this study have implications in at- the domain is densely populated by urban surfaces and, actu- mospheric chemistry as well. As the chemical transforma- ally, most of the corresponding 10km × 10km grid boxes re- tion, transport, diffusion and deposition of chemical species ally contain at least a few urban subgrid boxes. These might and aerosols are strongly linked to the meteorological (thus contribute to the impact on the whole grid box depending climate) conditions, the urban canopy induced changes cal- roughly on the percent urban coverage of the grid box. In- culated here may, in general, have impact on the air quality deed, Fig. 12 shows that if using only the 10km × 10km as well. This was a subject of a few studies focusing on Lon- resolution, the temperature impact is significantly larger over don and Seoul (Rigby and Toumi, 2008; Ryu et al., 2013a, these grid boxes than elsewhere. On the other hand, there is b). A future study will deal with the potential consequences still a relatively large impact over non-urbanized grid boxes on the air quality over central Europe as well.

Atmos. Chem. Phys., 14, 12393–12413, 2014 www.atmos-chem-phys.net/14/12393/2014/ P. Huszar et al.: Urban land-surface impact on climate 12411

Acknowledgements. This work has been funded by the Czech Deardoff, J. W.: Efficient prediction of ground surface temperature Science Foundation (GACR) project no. 13-19733P as well and moisture with inclusion of a layer of vegetation, J. Geophys. as by the project UHI funded in the framework of EC OP Res., 83, 1889–1903, 1978. NN (no. 3 CE292P3), the program PRVOUK – Environmental Dettwiller, J. and Changnon, S. A.: Possible urban effects on maxi- Research of the Charles University in Prague and, partially, mum daily rainfall at Paris, St. Louis and Chicago, J. Appl. Me- by the project UNCE 204020/2013. We further acknowledge teorol., 15, 518–519, 1976. the E-OBS data set from the EU-FP6 project ENSEMBLES Dickinson, R. E., Henderson-Sellers, A., and Kennedy, P.: Bio- (http://ensembles-eu.metoffice.com), the data providers in the sphere–atmosphere transfer scheme (BATS) version 1 as cou- ECA&D project (http://www.ecad.eu) and finally to ECMWF pled to the NCAR community climate model. Tech Rep, Na- ERA/Interim data sets. tional Center for Atmospheric Research Tech Note NCA R.TN- 387+STR, NCAR, Boulder, CO, 1993. Edited by: A. Hofzumahaus EEA: Corine Land Cover 2006 technical guidelines, EEA (Euro- pean Environment Agency), OPOCE (Office for Official Publi- cations of the European Communities), Copenhagen, 2012. Eliasson, I. and Holmer, B.: Urban heat island circulation in Gote- borg, Sweden, Theor. Appl. Climatol., 42, 187–196, 1990. Emanuel, K. A. and M. Zivkovic-Rothman, M.: Development and References evaluation of a convection scheme for use in climate models, J. Atmos. Sci., 56, 1766–1782, 1999. Angevine, W. M., White, A. B., Senff, C. J., Trainer, M., Fallmann, J., Emeis, S., and Suppan, P.: Mitigation of urban heat Banta, R. M., and Ayoub, M. A.: Urban-rural contrasts in mixing stress – a modelling case study for the area of Stuttgart, Die Erde, height and cloudiness over Nashville in 1999, J. Geophys. Res., 144, 202–216, doi:10.12854/erde-144-15, 2013. 108, 4092, doi:10.1029/2001JD001061, 2003. Farias, W. R. G., Pinto Jr., O., Naccarato, K. P., and Anthes, R. A.: A cumulus parameterization scheme utilizing a one- Pinto, I. R. C. A.: Anomalous lightning activity over the dimensional cloud model, Mon. Weather Rev., 105, 270–286, Metropolitan Region of São Paulo due to urban effects, Atmos. 1977. Res., 91, 485–490, 2009. Baklanov, A., Korsholm, U., Mahura, A., Petersen, C., and Feng, J.-M., Wang, Y.-L., Ma, Z.-G., and Liu, Y.-H.: Simulating Gross, A.: ENVIRO-HIRLAM: on-line coupled modelling of the regional impacts of urbanization and anthropogenic heat urban meteorology and air pollution, Adv. Sci. Res., 2, 41–46, release on climate across china, J. Climate, 25, 7187–7203, doi:10.5194/asr-2-41-2008, 2008. doi:10.1175/JCLI-D-11-00333.1, 2012. Basara, J. B., Hall Jr., P. K., Schroeder, A. J., Illston, B. G., Feng, J.-M., Wang, Y.-L., and Ma, Z.-G.: Long-term simulation of and Nemunaitis, K. L.: Diurnal cycle of the Oklahoma large-scale urbanization effect on the East Asian monsoon, Clim. City urban heat island, J. Geophys. Res., 113, D20109, Change, 1–13, doi:10.1007/s10584-013-0885-2, 2013. doi:10.1029/2008JD010311, 2008. Flagg, D. D. and Taylor, P. A.: Sensitivity of mesoscale model urban Block, A., Keuler, K., and Schaller, E.: Impacts of anthropogenic boundary layer meteorology to the scale of urban representation, heat on regional climate patterns, Geophys. Res. Lett., 31, Atmos. Chem. Phys., 11, 2951–2972, doi:10.5194/acp-11-2951- L12211, doi:10.1029/2004GL019852, 2004 2011, 2011. Bottyan, Z. and Unger, J.: A multiple linear statistical model for Fritsch, J. M. and Chappell, C. F.: Numerical prediction of con- estimating the mean maximum urban heat island, Theor. Appl. vectively driven mesoscale pressure systems, Part I: Convective Climatol., 75, 233–243, 2003. parameterization, J. Atmos. Sci., 37, 1722–1733, 1980. Changnon, S. A., Shealy, R. T., and Scott, R. W.: Precipitation Gaffin, S. R., Rosenzweig, C., Khanbilvardi, R., Parshall, L., Ma- Changes in Fall, Winter, and Spring Caused by St. Louis, J. Appl. hani, S., Glickman, H., Goldberg, R., Blake, R., Slosberg, R. B., Meteorol., 30, 126–134, 1991. and Hillel, D.: Variations in New York City’s urban heat island Chen, F., Kusaka, H., Bornstein, R., Ching, J., Grimmond, C. S. B., strength over time and space, Theor. Appl. Climatol., 94, 1–11, Grossman-Clarke, S., Loridan, T., Manning, K. W., Martilli, A., doi:10.1007/s00704-007-0368-3, 2008. Miao, S., Sailor, D., Salamanca, F. P., Taha, H., Tewari, M., Giannaros, T. M. and Melas, D.: Study of the urban heat Wang, X., Wyszogrodzki, A. A., and Zhang, C.: The integrated island in a coastal Mediterranean city: the case study WRF/urban modelling system: development, evaluation, and ap- of Thessaloniki, Greece, Atmos. Res., 118, 103–120, plications to urban environmental problems, Int. J. Climatol., 31, doi:10.1016/j.atmosres.2012.06.006, 2012. 273–288, 2011. Giannaros, T. M., Melas, D., Daglis, I. A., Keramitsoglou, I., and Cleugh, H. A. and Grimmond, C. S. B.: Modelling regional scale Kourtidis, K.: Numerical study of the urban heat island over surface energy exchanges and CBL growth in a heterogeneous, Athens (Greece) with the WRF model, Atmos. Environ., 73, urbanrural landscape, Bound.-Lay. Meteorol., 98, 1–31, 2001. 103–111, doi:10.1016/j.atmosenv.2013.02.055, 2013. Collier, C. G.: The impact of urban areas on weather, Q. J. Roy. Giorgi, F., Francisco, R., and Pal, J. S.: Effects of a sub-grid scale Meteor. Soc., 132, 1–25, 2006. topography and landuse scheme on surface climate and hydrol- Coquillat, S., Boussaton, M.-P., Buguet, M., Lambert, D., Rib- ogy. I. Effects of temperature and water vapor disaggregation, J. aud, J.-F., and Berthelot, A.: Lightning ground flash patterns Hydrometeorol., 4, 317–333, 200 over Paris area between 1992 and 2003: influence of pollution?, Giorgi, F., Coppola, E., Solmon, F., Mariotti, L., Sylla, M., Bi, X., Atmos. Res., 122, 77–92, doi:10.1016/j.atmosres.2012.10.032, Elguindi, N., Diro, G. T., Nair, V., Giuliani, G., Cozzini, S., Guet- 2013. www.atmos-chem-phys.net/14/12393/2014/ Atmos. Chem. Phys., 14, 12393–12413, 2014 12412 P. Huszar et al.: Urban land-surface impact on climate

tler, I., O’Brien, T. A., Tawfi, A. B., Shalaby, A., Zakey, A., spheric Research Tech Note NCAR/TN-420+STR, NCAR, Boul- Steiner, A., Stordal, F., Sloan, L., and Brankovic, C.: RegCM4: der, CO, 1996. Model description and preliminary tests over multiple CORDEX Klaic,´ Z. B., Nitis, T., Kos, I., and Moussiopoulos, N.: Modification domains, Clim. Res., 52, 7–29, 2012. of local winds due to hypothetical urbanization of the Zagreb GLC: Global Land Cover 2000 database. European Commission, surroundings, Meteorol. Atmos. Phys., 79, 1–12, 2002. Joint Research Centre, available at: http://bioval.jrc.ec.europa. Kłysik, K.: Spatial and seasonal distribution of anthropogenic heat eu/products/glc2000/glc2000.php (last access: 14 July 2014), emissions in Åodz, Poland, Atmos. Environ., 30, 3397–3404, 2000. 1995. Godowitch, J. M., Ching, J. K. S., and Clarke, J. F.: Evolution of the Kueppers, L. M., Snyder, M. A., Sloan, L. C., Cayan, D., Jin, J., nocturnal inversion layer at an urban and nonurban site, J. Appl. Kanamaru, H., Kanamitsu, M., Miller, N. L., Tyree, M., Du, H., Meteorol. Clim., 24, 791–804, 1985. and Weare, B.: Seasonal temperature response to land-use Grell, G.: Prognostic evaluation of assumptions used by cumulus change in the western United States, Global Planet. Change, 60, parameterizations, Mon. Weather Rev., 121, 764–787, 1993. 250–264, 2008. Grell, G., Dudhia, J., and Stauffer, D. R.: A description of the fifth Kusaka, H. and Kimura, F.: Coupling a single-layer urban canopy generation Penn State/NCAR Mesoscale Model (MM5), Na- model with a simple atmospheric model: impact on urban heat tional Center for Atmospheric Research Tech Note NCAR/TN- island simulation for an idealized case, J. Meteorol. Soc. Jpn., 398 + STR, NCAR, Boulder, CO, 1994. 82, 67–80, 2004. Grimmond, C. S. B. and Oke, T.: An evapotranspiration- Kusaka, H., Kondo, H., Kikegawa, Y., and Kimura, F.: A simple interception model for urban areas, Wafer Resour. Res. 27, 1739, singlelayer urban canopy model for atmospheric models: com- doi:10.1029/91WR00557, 1991. parison with multi-layer and slab models, Bound.-Lay. Meteo- Grimmond, C. S. B. and Oke, T. R.: Comparison of heat fluxes from rol., 101, 329–358, 2001. summertime observations in the suburbs of four North American Lee, S.-H. and Park, S.-U.: A vegetated urban canopy model for cities. J. Appl. Meteorol., 34, 873–889, 1995 meteorological and environmental modelling, Bound.-Lay. Me- Grimmond, C. S. B. and Oke, T. R.: Turbulent heat fluxes in ur- teorol., 126, 73–102, 2008. ban areas: Observations and a local-scale urban meteorologi- Lopes, A., Saraiva, J., and Alcoforado, M. J.: Urban boundary layer cal parameterization scheme (LUMPS), J. Appl. Meteorol., 41, wind speed reduction in summer due to urban growth and envi- 792–810, 2002. ronmental consequences in Lisbon, Environ. Modell. Softw., 54, Haeger-Eugensson, M. and Holmer, B.: Advection caused by the 611–623, doi:10.1016/j.atmosenv.2012.02.024, 2012. UHIC as a regulating factor on the nocturnal UHI, Int. J. Clima- Masson, V.: A physically-based scheme for the urban energy bud- tol., 19, 975–988, 1999. get in atmospheric models, Bound.-Lay. Meteorol., 94, 357–397, Haylock, M. R., Hofstra, N., Klein Tank, A. M. G., Klok, E. J., 2000. Jones, P. D., and New, M.: A European daily high-resolution Martilli, A: Numerical study of urban impact on boundary layer gridded dataset of surface temperature and precipitation. J. Geo- structure: sensitivity to wind speed, urban morphology, and rural phys. Res.-Atmos., 113, D20119, doi:10.1029/2008JD010201, soil moisture, J. Appl. Meteorol., 41, 1247–1266, 2002. 2008. Martilli, A., Clappier, A., and Rotach, M. W.: An urban surface Hidalgo, J., Masson, V., and Gimeno, L.: Scaling the Daytime Ur- exchange parameterization for mesoscale models, Bound.-Lay. ban Heat Island and Urban-Breeze Circulation, J. Appl. Meteo- Meteorol., 104, 261–304, 2002. rol. Clim., 49, 889–901, 2010. Miao, S., Chen, F., LeMone, A. M., Tewari, M., Li, Q., and Holtslag, A. A. M., Bruijn, E. I. F. D., and Pan, H.-L.: A high res- Wang, Y., 2009: An observational and modelling study of char- olution air mass transformation model for short-range weather acteristics of urban heat island and boundary layer structures in forecasting, Mon. Weather Rev., 118, 1561–1575, 1990. Beijing, J. Appl. Meteorol. Clim., 48, 484–501, 2009. Hou, A., Ni, G., Yang, H., and Lei, Z.: Numerical Analysis on the Nair, K. N., Freitas, E. D., Sanchez-Ccoyllo, O. R., Dias, M., Contribution of Urbanization to Wind Stilling: an Example over Dias, P. L. S., Andrade, M. F., and Massambani, O.: Dynamics of the Greater Beijing Metropolitan Area, J. Appl. Meteorol. Clim., urban boundary layer over Sao Paulo associated with mesoscale 52, 1105–1115, doi:10.1175/JAMC-D-12-013.1, 2013. processes, Meteorol. Atmos. Phys., 86, 87–98, 2004. Jauregui, E.: Influence of a large urban park on temperature and Oke, T. R.: The energetic basis of the urban heat island, Q. J. Roy. convective precipitation in a tropical city, Energ. Buildings, 15, Meteor. Soc., 108, 1–24, doi:10.1002/qj.49710845502, 1982. 457–463, 1991. Oke, T. R.: Boundary Layer Climates, Methuen and Co. Ltd, Lon- Kastner-Klein, P., Fedorovich, E., and Rotach, M. W.: A wind tun- don, 2nd Edn., 1987. nel study of organised and turbulent air motions in urban street Oleson, K. W., Bonan, J. B., Feddema, J., Vertenstein, M., and canyons, J. Wind Eng. Ind. Aerod., 89, 849–861, 2001. Grimmond, C. S. B.: An urban parameterization for a global cli- Kaufmann, R. K., Seto, K. C., Schneider, A., Liu, Z., Zhou, L., mate model, Part I: Formulation and evaluation of two cities, J. and Wang, W.: Climate response to rapid urban growth: evi- Appl. Meteorol. Clim., 47, 1038–1060, 2008. dence of a human-induced precipitation deficit, J. Climate, 20, Pal, J. S., Small, E. E., and Eltahir, E. A. B.: Simulation of regional- 2299–2306, 2007. scale water and energy budgets: Representation of subgrid cloud Kiehl, J., Hack, J., Bonan, G., Boville, B., Breigleb, B., and precipitation processes within RegCM, J. Geophys. Res.- Williamson, D., and Rasch, P.: Description of the NCAR Com- Atmos., 105, 29579–29594, 2000. munity Climate Model (CCM3). National Center for Atmo- Pichierri, M., Bonafoni, S., and Biondi, R.: Satellite air temperature estimation for monitoring the canopy layer heat

Atmos. Chem. Phys., 14, 12393–12413, 2014 www.atmos-chem-phys.net/14/12393/2014/ P. Huszar et al.: Urban land-surface impact on climate 12413

island of Milan, Remote Sens. Environ., 127, 130–138, Simmons, A., Uppala, S., Dee, D., and Kobayashi, S.: ERA- doi:10.1016/j.rse.2012.08.025, 2012. Interim: New ECMWF Reanalysis Products from 1989 Onwards, Pigeon, G., Legain, D., Durand, P., and Masson, V.: Anthropogenic Newsletter 110, Winter 2006/07, ECMWF, Reading, 2007. heat release in an old European agglomeration (Toulouse, Skamarock, W. C., Klemp, J. B., Dudhia, J., Gill, D. O., France), Int. J. Climatol., 27, 1969–1981, 2007. Barker, D. M., Duda, M., Huang, X.-Y., Wang, W., and Pow- Piringer, M.: Exploring the urban boundary layer by sodar and teth- ers, J. G.: A Description of the Advanced Research WRF Ver- ersonde, Phys. Chem. Earth Pt. B., 26, 881–885, 2001. sion 3 NCAR Technical Note, National Center for Atmospheric Pongrácz, R., Bartholy, J., and Dezso,˝ Z.: Application of re- Research, Boulder CO, USA, doi:10.5065/D68S4MVH, 2008. motely sensed thermal information to urban climatology of Struzewska, J. and Kaminski, J. W.: Impact of urban parameteriza- Central European cities, Phys. Chem. Earth, 35, 95–99, tion on high resolution air quality forecast with the GEM – AQ doi:10.1016/j.pce.2010.03.004, 2010. model, Atmos. Chem. Phys., 12, 10387–10404, doi:10.5194/acp- Ratti, C., Di Sabatino, S., Britte, R., Brown, M., Caton, F., and 12-10387-2012, 2012. Burian, S.: Analysis of 3-D urban databases with respect to pol- Svoma, B. M. and Brazel, A.: Urban effects on the diurnal temper- lution dispersion for a number of European and American cities, ature cycle in Phoenix, Arizona, Clim. Res., 41, 21–29, 2010. Water Air Soil Poll., 2, 459–469, 2001. Taha, H.: Urban climates and heat islands: Albedo, evapotranspi- Reid, C. E., O’Neill, M. S., Gronlund, C. J., Brines, S. J., ration, and anthropogenic heat, Energy. Buildings, 2, 99–103, Brown, D. G., Diez-Roux, A. V., and Schwartz, J.: Mapping 1997. community determinants of heat vulnerability, Environ. Health Tanaka, S., Takeda, H., Adachi, T., and Tsuchiya, T.: Architectural Persp., 117, 1730–1736, doi:10.1289/ehp.0900683, 2009. Environmental Engineering, Inoue Co., Ltd., Tokyo, 301 pp., Richards, K.: Observation and simulation of dew in rural and urban 2013 (in Japanese). environments, Prog. Phys. Geog., 28, 76–94, 2004. Tawfik, A. B. and Steiner, A. L.: The role of soil ice in Richards, K. and Oke, T. R.: Validation and results of a scale model land–atmosphere coupling over the United States: a soil mois- of dew deposition in urban environments, Int. J. Climatol., 22, ture precipitation winter feedback mechanism. J. Geophys. Res., 1915–1933, 2002. 116, D02113, doi:10.1029/2010JD014333, 2011. Rigby, M. and Toumi, R.: London air pollution climatol- Trusilova, K., Jung, M., Churkina, G., Karstens, U., Heimann, M., ogy: indirect evidence for urban boundary layer height and and Claussen, M.: Urbanization Impacts on the Climate in Eu- wind speed enhancement, Atmos. Environ., 42, 4932–4947, rope: Numerical Experiments by the PSU–NCAR Mesoscale doi:10.1016/j.atmosenv.2008.02.031, 2008. Model (MM5), J. Appl. Meteorol. Clim., 47, 1442–1455, Rotach, M. W., Richner, H., Vogt, R., Christen, A., Parlow, E., doi:10.1175/2007JAMC1624.1, 2008. Bernhofer, C., Batchvarova, E., Clappier, A., Roulet, Y.-A., Unger, J., Savic,´ S., and Gál, T.: Modelling of the annual mean Feddersen, B., Schatzmann, M., Batchvarova, E., Gryning, S.- urban heat island pattern for planning of representative ur- E., Mayer, H., Martucci, G., Mitev, V., Oke, T. R., Roth, M., ban climate station network, Adv. Meteorol., 2011, ID398613, Ruffieux, D., Salmond, J. A., and Voogt, J. A.: BUBBLE-an ur- doi:10.1155/2011/398613, 2011. ban boundary layer project, Theor. Appl. Climatol., 81, 231–261, Vautard, R., Cattiaux, J., Yiou, P., Thepaut, J., and Ciais, P.: 2005. Northern Hemisphere atmospheric stilling partly attributed to Roth, M.: Review of atmospheric turbulence over cities, Q. J. Roy. an increase in surface roughness, Nat. Geosci., 3, 756–761, Meteor. Soc., 126, 941–990, 2000. doi:10.1038/NGEO979, 2010. Rozoff, C. M., Cotton, W. R., and Adegoke, J. O.: Simulation of Wang, J., Feng, J., Yan, Z., Hu, Y., and Jia, G.: Nested high- St. Louis, Missouri, land use impacts on thunderstorms, J. Appl. resolution modeling of the impact of urbanization on regional Meteorol., 42, 716–738, 2003. climate in three vast urban agglomerations in China, J. Geophys. Ryu, Y.-H., Baik, J.-J., Kwak, K.-H., Kim, S., and Moon, N.: Im- Res., 117, D21103, doi:10.1029/2012JD018226, 2012. pacts of urban land-surface forcing on ozone air quality in the Wouters, H., De Ridder, K., Demuzere, M., Lauwaet, D., and Seoul metropolitan area, Atmos. Chem. Phys., 13, 2177–2194, van Lipzig, N. P. M.: The diurnal evolution of the urban heat doi:10.5194/acp-13-2177-2013, 2013a. island of Paris: a model-based case study during Summer 2006, Ryu, Y.-H., Baik, J.-J., and Lee, S.-H.: Effects of anthropogenic heat Atmos. Chem. Phys., 13, 8525–8541, doi:10.5194/acp-13-8525- on ozone air quality in a megacity, Atmos. Environ., 80, 20–30, 2013, 2013. doi:10.1016/j.atmosenv.2013.07.053, 2013b. Yang, X., Hou, Y., and Chen, B.: Observed surface warming in- Sailor, D. J. and Lu, L.: A top-down methodology for developing duced by urbanization in east China, J. Geophys. Res., 116, diurnal and seasonal anthropogenic heating profiles for urban ar- D14113, doi:10.1029/2010JD015452, 2011. eas, Atmos. Environ., 38, 2737–2748, 2004. Zhang, H., Gao, X., and Li, Y.: Climate impacts of land-use change Schaldach, R. and Alcamo, J.: Simulating the effects of urbaniza- in China and its uncertainty in a global model simulation, Clim. tion, afforestation and cropland abandonment on a regional car- Dynam., 32, 473–494, 2009. bon balance: a case study for Central Germany, Reg. Environ. Zhou, Y. and Ren, G.: Change in extreme temperature event Change, 7, 137–148, doi:10.1007/s10113-007-0034-4, 2007. frequency over mainland China, 1961–2008, Clim. Res., 50, Shepherd, J. M., Pierce, H., and Negri, A. J.: Rainfall modification 125–139, 2011. by major urban areas: observations from spaceborne rain radar on the TRMM satellite, J. Appl. Meteorol., 41, 689–701, 2002.

www.atmos-chem-phys.net/14/12393/2014/ Atmos. Chem. Phys., 14, 12393–12413, 2014